async/await in C# — .Result Deadlocks Your UI Thread

Every production .NET app eventually hits the same wall: a database query takes 200ms, an external API call takes a second, a file upload blocks the thread — and suddenly your server is choking on requests it should handle easily. Thread-per-request models waste memory and CPU context-switching on work that's just sitting idle waiting for I/O. In high-traffic systems, that's not just inefficient — it's a scalability killer. async/await is the reason modern ASP.NET Core can handle tens of thousands of concurrent requests on a handful of threads.

Before async/await arrived in C# 5, developers juggled callbacks, BeginInvoke/EndInvoke patterns, and event-based async patterns (EAP) — all of which produced code that was brittle, hard to read, and nearly impossible to debug. async/await solved this by letting you write asynchronous code that looks almost identical to synchronous code, while the compiler does the heavy lifting behind the scenes, transforming your method into a state machine.

By the end you'll understand not just the syntax but why async/await works the way it does — including what the compiler actually generates, how the SynchronizationContext interacts with your awaits, when ConfigureAwait(false) is mandatory, how to avoid the deadlock that bites almost every developer once, and how to squeeze maximum performance out of async patterns in real production scenarios.

What is async and await in C#?

async and await are keywords that enable non-blocking asynchronous programming. When you mark a method with async, the compiler transforms it into a state machine. The await keyword suspends the method until the awaited operation completes — but without blocking the calling thread. This is fundamentally different from synchronous code where a thread sits idle waiting for I/O.

The key insight: async methods return a Task or Task<T> immediately to the caller. The actual work continues asynchronously. When the operation completes, the continuation runs on the captured context (or thread pool pool if ConfigureAwait(false) is used). This allows a single thread to handle many concurrent operations, drastically improving scalability.

For example, an ASP.NET Core endpoint can handle thousands of concurrent database queries with only a handful of threads. When one query awaits a database call, that thread is released back to the thread pool to serve another request. Once the database responds, a thread picks up the continuation. That's the scalability magic.

using System;
using System.Net.Http;
using System.Threading.Tasks;

namespace io.thecodeforge.csharp
{
    public class AsyncExample
    {
        private static readonly HttpClient httpClient = new HttpClient();

        public async Task<string> FetchDataAsync(string url)
        {
            Console.WriteLine("Starting fetch");
            string result = await httpClient.GetStringAsync(url);
            Console.WriteLine("Fetch completed");
            return result;
        }

        // Synchronous version for comparison
        public string FetchDataSync(string url)
        {
            Console.WriteLine("Starting fetch (sync)");
            // Blocks thread until complete
            string result = httpClient.GetStringAsync(url).Result;
            Console.WriteLine("Fetch completed (sync)");
            return result;
        }
    }
}

The State Machine — What the Compiler Generates

When the compiler encounters an async method, it generates a state machine structure. The method is rewritten as a method that returns a Task (or Task<T>) and creates an instance of the state machine.

The MoveNext() method is called to advance the state machine. Each await splits the method into a 'before' and 'after' part. When the awaited operation completes, it calls MoveNext() again on the captured state machine.

This transformation is why async methods can 'pause' without blocking threads. The method returns the Task to the caller immediately. The state machine lives on the heap, not the stack. When the operation completes, a thread (usually from the thread pool) calls MoveNext() to resume the method.

// Original async method:
public async Task<int> GetDataAsync()
{
    Console.WriteLine("Start");
    int data = await GetRemoteDataAsync();
    Console.WriteLine($"Got: {data}");
    return data * 2;
}

// Simplified decompiled state machine (what compiler generates):
private sealed class GetDataAsyncStateMachine : IAsyncStateMachine
{
    public int state;
    public AsyncTaskMethodBuilder<int> builder;
    public TaskAwaiter<int> awaiter;
    private int data;

    void MoveNext()
    {
        int result;
        try
        {
            if (state == 0)
            {
                Console.WriteLine("Start");
                var awaiter = GetRemoteDataAsync().GetAwaiter();
                if (!awaiter.IsCompleted)
                {
                    state = 1;
                    builder.AwaitUnsafeOnCompleted(ref awaiter, ref this);
                    return;
                }
                data = awaiter.GetResult();
            }
            else if (state == 1)
            {
                data = awaiter.GetResult();
            }
            Console.WriteLine($"Got: {data}");
            result = data * 2;
            builder.SetResult(result);
        }
        catch (Exception ex)
        {
            builder.SetException(ex);
        }
    }
}

Async State Machine Visual Architecture

The async state machine is not just a theoretical concept — it is a concrete struct generated by the compiler for every async method. Understanding its lifecycle helps you write more performant async code and diagnose allocation issues in production.

The diagram below illustrates the journey of an async method call: the caller invokes the async method, the compiler-emitted state machine is allocated, and execution proceeds through the method until hitting an await on an incomplete operation. At that point, the state machine is boxed and a continuation is registered with the awaiter. When the operation completes, the continuation (MoveNext) is invoked, often via the captured SynchronizationContext, and the state machine advances to the next state.

In hot paths, this allocation and boxing overhead matters. Each async call that completes asynchronously (i.e., the await sees an incomplete task) triggers a heap allocation for the state machine. If the task completes synchronously, the state machine is typically not allocated, thanks to a compiler optimisation. Knowing this, you can profile your code to identify unnecessary allocations by checking which async methods frequently await incomplete tasks.

Context Capture — Visual Flow Diagram

The SynchronizationContext capture occurs at each await point unless ConfigureAwait(false) is specified. The captured context determines where the continuation runs — on the original thread (UI, classic ASP.NET) or on a thread pool thread. This section visualises the flow for a UI application where the default context is the UI thread.

In WPF, the DispatcherSynchronizationContext posts continuations to the UI thread's message pump. When the async method hits an await on an incomplete task, the current SynchronizationContext is captured (unless ConfigureAwait(false) tells it not to). The continuation is posted as a message to the UI thread queue. When the awaited operation completes, that message is processed and the state machine resumes on the UI thread. This is essential for updating UI controls, but it also creates the deadlock risk when the UI thread is blocked.

The diagram below shows a healthy flow: the UI thread awaits, releases, and later the continuation returns to the UI thread to update the UI. Then it shows the deadlock scenario: the UI thread is blocked by .Result, so the continuation message cannot be processed, causing a circular wait.

SynchronizationContext and ConfigureAwait(false)

Every await captures the current SynchronizationContext (or TaskScheduler) unless configured otherwise. When the awaited operation completes, the continuation runs on that captured context.

In UI applications (WPF, WinForms, MAUI), the SynchronizationContext posts work to the UI thread. In ASP.NET (pre-Core), the context ensures the continuation runs on the same request thread (with HttpContext). In ASP.NET Core, there is NO SynchronizationContext by default (improvement).

ConfigureAwait(false) tells the awaiter NOT to capture the current context. The continuation can run on any thread pool thread. This: (a) avoids deadlocks (continuation doesn't need the captured thread), (b) improves performance (avoids unnecessary thread switches), (c) should be used in library code that doesn't need original context.

using System;
using System.Threading.Tasks;
using System.Windows.Forms;

public class ConfigureAwaitDemo
{
    // BAD: This deadlocks in UI app
    public async Task<string> GetDataAndBlockDeadlock()
    {
        // Captures UI SynchronizationContext
        var data = await GetRemoteDataAsync();
        return data;
    }

    // GOOD: ConfigureAwait(false) prevents deadlock and improves performance
    public async Task<string> GetDataConfigured()
    {
        // Does NOT capture context — runs continuation on thread pool
        var data = await GetRemoteDataAsync().ConfigureAwait(false);
        return data;
    }

    // Event handler — must capture context to update UI
    private async void Button_Click(object sender, EventArgs e)
    {\n        // Omit ConfigureAwait(false) — need UI thread after await\n        var data = await GetRemoteDataAsync();\n        textBox1.Text = data; // Requires UI thread\n    }

    // Library code — always use ConfigureAwait(false)
    public async Task<string> FetchUserDataAsync(int userId)
    {
        var user = await db.Users.FindAsync(userId).ConfigureAwait(false);
        var orders = await db.Orders.Where(o => o.UserId == userId)
                                    .ToListAsync()
                                    .ConfigureAwait(false);
        return $"{user.Name} has {orders.Count} orders";
    }
}

Async Void — The Only Time You Should Use It (And Why It's Dangerous)

async void methods have different error handling semantics than async Task. Exceptions thrown in async void are raised on the synchronization context (typically crashing the process). They cannot be caught with try-catch outside the method. They also cannot be awaited, so callers have no way to know when they complete.

The only legitimate use of async void is for event handlers (async void Button_Click(object sender, EventArgs e)). The event dispatcher expects a void return type; you cannot change it to Task. For all other cases, return Task (for no result) or Task (for result).

If you must use async void (event handlers), wrap the body in try-catch and log exceptions to avoid process crashes.

using System;
using System.Threading.Tasks;

public class AsyncVoidDangers
{
    // DANGEROUS: This will crash the process if an exception is thrown
    public async void ProcessDataAsync()
    {
        await Task.Delay(100);
        throw new InvalidOperationException("Crash!"); // Process terminates
    }

    // GOOD: Return Task — exception caught by caller
    public async Task ProcessDataGoodAsync()
    {
        await Task.Delay(100);
        throw new InvalidOperationException("Handled by caller");
    }

    // SAFER EVENT HANDLER: Wrap in try-catch, log exceptions
    private async void OnButtonClick(object sender, EventArgs e)
    {
        try
        {
            await DoWorkAsync();
        }
        catch (Exception ex)
        {
            // Log to file, send to telemetry, show user dialog
            Console.WriteLine($"Error in event handler: {ex}");
        }
    }

    private async Task DoWorkAsync()
    {
        await Task.Delay(100);
        // Simulate work
    }
}

Async Patterns: Task.WhenAll, Task.WhenAny, and Structured Concurrency

Real-world async code often involves multiple independent operations. You might need to fetch data from three APIs simultaneously and wait for all of them, or start several tasks and take action when the first one completes. That's where Task.WhenAll and Task.WhenAny come in.

Task.WhenAll takes a collection of tasks and returns a single task that completes when all of them have completed. If any task faults, the returned task faults with an AggregateException containing all exceptions. Use await Task.WhenAll(tasks) for fan-out scenarios where tasks are independent.

Task.WhenAny completes when the first task completes. It's useful for race conditions, timeout wrappers, or load-balancing across redundant endpoints.

Important: when using WhenAll with a large number of tasks, consider batching with SemaphoreSlim to control concurrency and avoid memory pressure from too many in-flight operations. Never fire-and-forget tasks unless you have explicit exception handling.

using System;
using System.Collections.Generic;
using System.Net.Http;
using System.Threading.Tasks;

namespace io.thecodeforge.csharp
{
    public class ConcurrencyPatterns
    {
        private static readonly HttpClient client = new HttpClient();

        // Fetch multiple URLs concurrently
        public async Task<List<string>> FetchAllAsync(IEnumerable<string> urls)
        {
            var tasks = new List<Task<string>>();
            foreach (var url in urls)
            {
                tasks.Add(client.GetStringAsync(url));
            }

            // WhenAll returns Task<string[]> – you can await directly
            string[] results = await Task.WhenAll(tasks);
            return new List<string>(results);
        }

        // First completed wins
        public async Task<string> FirstRespondingAsync(string primaryUrl, string fallbackUrl)
        {\n            var primaryTask = client.GetStringAsync(primaryUrl);\n            var fallbackTask = client.GetStringAsync(fallbackUrl);\n\n            Task<string> completedTask = await Task.WhenAny(primaryTask, fallbackTask);\n            return await completedTask;\n        }
    }
}

Task vs ValueTask — Decision Matrix and Performance Guide

Task and ValueTask both represent asynchronous operations, but they differ in allocation behaviour. Task is a reference type — each async method call that returns a Task allocates a new Task object (and a state machine if the method is async and the await is on an incomplete task). ValueTask is a value type, introduced to reduce memory allocations when the result is frequently available synchronously.

The decision matrix below clarifies when to choose each type.

using System;
using System.Threading.Tasks;

public class ValueTaskExample
{
    // Good candidate for ValueTask: often returns cached result
    private string cachedData;
    private bool cacheValid;

    public async ValueTask<string> GetDataAsync()
    {
        if (cacheValid)
            return cachedData;  // synchronous fast path — no allocation
        
        cachedData = await FetchFromDatabaseAsync();
        cacheValid = true;
        return cachedData;
    }

    // Stay with Task<T> if result is rarely synchronous or caller awaits multiple times
    public async Task<string> FetchFromDatabaseAsync()
    {
        // simulate async I/O
        await Task.Delay(100);
        return "database result";
    }
}

Async Best Practices Cheat Sheet

Below is a concise table summarising the most important rules for writing reliable, performant async code in C#. Use this as a quick reference during code reviews or when designing new async APIs.

// Example of good practices in one file
using System;
using System.Threading.Tasks;

public class BestPracticesDemo
{
    // Rule 1: I/O — always async
    public async Task<string> GetDataAsync()
    {
        using var client = new HttpClient();
        return await client.GetStringAsync("url").ConfigureAwait(false);
    }

    // Rule 2: CPU-bound — use Task.Run, not async
    public Task<int> ComputeAsync(int[] data)
    {
        return Task.Run(() =>
        {
            // CPU-heavy work
            return Array.IndexOf(data, 42);
        });
    }

    // Rule 3: Never block on async — avoid .Result and .Wait()
    // Good: await all the way

    // Rule 4: async void only for event handlers
    private async void OnClick(object sender, EventArgs e)
    {\n        try { await DoWorkAsync(); }
        catch (Exception ex) { Log(ex); }
    }

    // Rule 5: ConfigureAwait(false) in library code
    public async Task<string> LibraryMethodAsync()
    {
        await Task.Delay(10).ConfigureAwait(false);
        return "done";
    }

    private Task DoWorkAsync() => Task.CompletedTask;
    private void Log(Exception ex) { }
}

Don't Block, Await Instead — The Thread Pool Tax

Blocking on async code is the fastest way to burn threads and crater throughput. .Result, .Wait(), and Task.WaitAll() are the devil's tools. They cause thread pool starvation by grabbing a thread, parking it, and praying the async operation finishes before the thread pool's injection heuristic kicks in. When it doesn't, you deadlock or degrade to single-digit request rates.

The fix is brutal honesty: mark the calling method async and await. If you're in a constructor, capture the task and restructure. If you're in an event handler, make it async Task and handle exceptions properly. ConfigureAwait(false) isn't a magic bullet for blocking — it avoids capturing the SynchronizationContext, but a blocking call still hogs a thread.

A production incident taught me this: 200 concurrent requests hitting a blocking Task.Result on an HTTP call caused 15-second response times. Switching to await dropped latency by 90% and freed 40 threads. Blocking is gambling with your thread pool.

// io.thecodeforge — csharp tutorial

using System;
using System.Net.Http;
using System.Threading.Tasks;

public class MetricsService
{
    private readonly HttpClient _client;

    public MetricsService(HttpClient client) => _client = client;

    // ❌ Never do this — blocks thread and may deadlock
    public string GetMetricsBlocking()
    {
        return _client.GetStringAsync("https://metrics.internal").Result;
    }

    // ✅ Do this — releases thread while I/O completes
    public async Task<string> GetMetricsAsync()
    {
        return await _client.GetStringAsync("https://metrics.internal");
    }
}

// Output (n/a — compiles, but runtime behavior differs)
// Blocking: thread blocks for ~500ms, uses 1 thread per request.
// Async: releases thread during I/O, uses 0 threads while waiting.

async/await vs ContinueWith — The Lambda Trap

Before async/await, we had ContinueWith to chain tasks. It works, but it's a footgun for exception handling and context capture. await gives you the compiler's state machine, which unwraps AggregateException, captures SynchronizationContext, and reads like synchronous code. ContinueWith dumps you into a lambda, leaves exceptions wrapped, and forces manual context management.

Look at the code below. The ContinueWith version swallows the exception unless you explicitly check task.Exception. The await version throws naturally — and the catch block works as expected. In a real incident, a junior wrote ContinueWith without error handling; the service silently failed for three hours before we discovered the unobserved task exception.

When should you still use ContinueWith? Almost never. Only when you're building a custom task scheduler or need fine-grained TaskContinuationOptions (e.g., OnlyOnRanToCompletion). For 99% of code, await wins. Your cognitive load is lower, your exception flow is predictable, and your teammates won't have to decode nested lambdas during an outage.

// io.thecodeforge — csharp tutorial

using System;
using System.Net.Http;
using System.Threading.Tasks;

public class ContinuationDemo
{
    private static readonly HttpClient Client = new();

    public Task<string> BadWay()
    {
        return Client.GetStringAsync("https://api.payments")
            .ContinueWith(task =>
            {
                // Exception is wrapped in AggregateException
                return task.Result.ToUpper(); // throws if failed
            });
    }

    public async Task<string> GoodWay()
    {
        try
        {
            var data = await Client.GetStringAsync("https://api.payments");
            return data.ToUpper();
        }
        catch (HttpRequestException ex)
        {
            throw new PaymentServiceException("Payment API call failed", ex);
        }
    }
}

// Output (n/a — demonstrates exception handling difference)
// BadWay: Exception hidden in AggregateException, unobserved = swallowed.
// GoodWay: Exception thrown cleanly, catch block fires immediately.

Handle Asynchronous Exceptions — No Silent Failures

Async exceptions behave like synchronous ones — but only if you await. An unobserved Task that throws will eventually trigger TaskScheduler.UnobservedTaskException and tear down your process in .NET Core+. Silent failures in async code are worse than synchronous crashes because they surface minutes later as a poisoned app pool or hung requests.

The golden rule: always await or attach a continuation. If you fire-and-forget, the exception disappears into the void. In production, I've traced a 15-minute outage to an async void event handler that threw NullReferenceException — nobody knew until the next deploy recycled the process. async void cannot be awaited, so exceptions escape to the SynchronizationContext.Post handler, which crashes the app domain.

Pattern your error handling: wrap await calls in try/catch for known failures (HTTP 500s, timeouts). For unhandled exceptions, hook TaskScheduler.UnobservedTaskException in your startup for telemetry. Never hide the stack trace — it's your only breadcrumb. And if you must fire-and-forget, log the exception immediately inside the lambda.

// io.thecodeforge — csharp tutorial

using System;
using System.Threading.Tasks;

public class PaymentHandler
{
    // ❌ Async void — exception crashes the process
    public async void FireAndForget()
    {
        await Task.Delay(100);
        throw new InvalidOperationException("DB connection lost");
    }

    // ✅ Async Task — caller can catch
    public async Task SafeCall()
    {
        try
        {
            await SimulatePayment();
        }
        catch (PaymentException ex)
        {
            Logger.Log($"Payment failed: {ex.Message}");
            throw; // re-throw if caller should also know
        }
    }

    private async Task SimulatePayment()
    {
        await Task.Delay(50);
        throw new PaymentException("Gateway timeout");
    }
}

public class PaymentException : Exception
{
    public PaymentException(string message) : base(message) { }
}

// Output (runtime exception propagation):
// FireAndForget: unhandled exception → process crash (or hang w/ unobserved)
// SafeCall: exception caught and logged, stack trace preserved

Why async/await Is Preferred — Lower Complexity, Higher Safety

async/await is preferred over raw Task continuations because it preserves the logical control flow of synchronous code. When you use ContinueWith, each lambda creates a closure that captures variables by reference, leading to race conditions and unintended side effects. async/await eliminates these issues by letting the compiler generate a safe state machine that manages context and exception propagation automatically. The key advantage is structured concurrency: the await keyword signals a suspension point that the runtime handles without blocking threads. This avoids the thread pool tax where blocked threads waste memory and CPU. Furthermore, async/await integrates seamlessly with using blocks and try-catch, ensuring resources are disposed and exceptions bubble correctly. In production, preferring async/await means your code reads like a sequential narrative, reducing cognitive load for reviewers and preventing the silent failures common in manual continuation patterns.

// io.thecodeforge — csharp tutorial

// Shows why async/await beats ContinueWith
public async Task<string> FetchDataAsync()
{
    // Bad: ContinueWith creates closure race conditions
    var bad = httpClient.GetStringAsync(url)
        .ContinueWith(t => Process(t.Result));

    // Good: async/await preserves control flow safely
    var data = await httpClient.GetStringAsync(url);
    return Process(data);
}

private string Process(string input) => input.ToUpper();

Apply await Expressions to Tasks Efficiently — Minimize Suspension Overhead

Every await suspension has a measurable cost: the compiler must save and restore the state machine context, which adds CPU cycles and memory allocations. To apply await efficiently, avoid unnecessary suspensions by awaiting only when a result is genuinely needed to proceed. Batch independent tasks with Task.WhenAll before awaiting them, so the state machine suspends once for all parallel work rather than sequentially. Additionally, use ValueTask for hot paths where the result is often synchronous, preventing heap allocations for the task object. When the awaited task completes before the await, the state machine avoids suspension entirely — structure your code to keep awaited tasks hot. Finally, never await inside tight loops that don't depend on the previous iteration's result; instead, collect tasks in a list and await them outside the loop. This pattern reduces state machine transitions, cutting latency in high-throughput systems.

// io.thecodeforge — csharp tutorial

// Inefficient: sequential awaits on independent tasks
for (int i = 0; i < 10; i++)
{
    var result = await FetchItemAsync(i); // 10 suspensions
}

// Efficient: batch then await once
var tasks = Enumerable.Range(0, 10)
    .Select(i => FetchItemAsync(i)).ToArray();
var results = await Task.WhenAll(tasks); // 1 suspension

ConfigureAwait with ConfigureAwaitOptions (C# 13/.NET 9)

In C# 13 and .NET 9, the ConfigureAwait method has been enhanced with ConfigureAwaitOptions to provide finer-grained control over how awaiting tasks interacts with the synchronization context. Previously, ConfigureAwait(false) was the primary way to avoid deadlocks by not capturing the current context. Now, you can specify options such as SuppressContext, ForceYielding, and ContinueOnCapturedContext explicitly.

ConfigureAwaitOptions.SuppressContext is equivalent to ConfigureAwait(false) — it suppresses the capture of the current synchronization context. ForceYielding ensures that the continuation is always scheduled as a new task, even if the operation completes synchronously, which can help avoid stack overflows in recursive patterns. ContinueOnCapturedContext explicitly forces capturing the context, which is the default behavior.

Here's an example of using ConfigureAwaitOptions:

public async Task<string> FetchDataAsync()
{
    using var httpClient = new HttpClient();
    var result = await httpClient.GetStringAsync("https://api.example.com/data")
        .ConfigureAwait(ConfigureAwaitOptions.SuppressContext);
    return result;
}

This is particularly useful in library code where you want to avoid deadlocking the calling UI thread. The new options also allow you to combine flags, e.g., ConfigureAwaitOptions.SuppressContext | ConfigureAwaitOptions.ForceYielding. However, use ForceYielding sparingly as it adds overhead.

In production, adopt ConfigureAwaitOptions.SuppressContext in all library methods that are not UI-bound. For application code that interacts with UI, you may want to keep context capture for certain continuations. The new API makes the intent clearer and reduces the chance of accidentally omitting false.

using System;
using System.Net.Http;
using System.Threading.Tasks;

public class DataService
{
    private readonly HttpClient _httpClient = new();

    public async Task<string> GetDataAsync()
    {
        // Using ConfigureAwaitOptions.SuppressContext to avoid deadlocks
        var result = await _httpClient.GetStringAsync("https://api.example.com/data")
            .ConfigureAwait(ConfigureAwaitOptions.SuppressContext);
        return result;
    }

    public async Task ProcessDataAsync()
    {
        // ForceYielding can help in deep recursion scenarios
        await Task.Yield();
        // ... processing
    }
}

Async Disposal with IAsyncDisposable and await using

Starting with C# 8.0, .NET introduced IAsyncDisposable to enable asynchronous cleanup of resources, such as closing network connections or flushing file streams. The await using statement allows you to asynchronously dispose of objects that implement IAsyncDisposable. This is crucial for scenarios where disposal involves I/O operations that should not block the calling thread.

To implement IAsyncDisposable, you need to provide a DisposeAsync method that returns a ValueTask. Here's an example:

public class AsyncResource : IAsyncDisposable
{
    private bool _disposed;
public async ValueTask DisposeAsync()
    {
        if (!_disposed)
        {
            _disposed = true;
            // Simulate async cleanup
            await Task.Delay(100);
        }
    }
}

You can then use await using to ensure proper disposal:

public async Task UseResourceAsync()
{
    await using var resource = new AsyncResource();
    // Use the resource
} // DisposeAsync is called automatically here

It's important to note that IAsyncDisposable and IDisposable are independent. A class can implement both if it has both synchronous and asynchronous cleanup paths. In practice, prefer IAsyncDisposable when disposal involves async operations.

In production, always implement IAsyncDisposable for classes that hold async resources like database connections, HTTP clients, or file streams. Use await using to guarantee cleanup even if exceptions occur. Be aware that the await using pattern is not supported in all contexts (e.g., older C# versions), so check your target framework.

using System;
using System.Threading.Tasks;

public class AsyncDatabaseConnection : IAsyncDisposable
{
    private bool _disposed;

    public async ValueTask DisposeAsync()
    {
        if (!_disposed)
        {
            _disposed = true;
            // Asynchronously close the connection
            await Task.Delay(50); // Simulate closing
            Console.WriteLine("Connection closed asynchronously.");
        }
    }
}

public class DataProcessor
{
    public async Task ProcessAsync()
    {
        await using var connection = new AsyncDatabaseConnection();
        // Use connection
        await Task.Delay(10);
    } // DisposeAsync called automatically
}

Parallel Async with Task.WhenAll and Task.WhenEach (.NET 9)

Running multiple asynchronous operations concurrently is a common need. Task.WhenAll has been the go-to method for awaiting a collection of tasks and collecting their results. However, it has limitations: it requires all tasks to complete before you can process any results, and it doesn't support streaming or early cancellation of individual tasks.

In .NET 9, Task.WhenEach was introduced to address these issues. It allows you to iterate over tasks as they complete, processing results one by one. This is especially useful for scenarios like making multiple HTTP requests and handling responses as they arrive, rather than waiting for all to finish.

Here's an example comparing both:

// Task.WhenAll - wait for all
var tasks = Enumerable.Range(1, 5).Select(i => FetchDataAsync(i));
var results = await Task.WhenAll(tasks);
foreach (var result in results)
{
    Console.WriteLine(result);
}
// Task.WhenEach - process as completed (.NET 9)
await foreach (var task in Task.WhenEach(tasks))
{
    var result = await task;
    Console.WriteLine(result);
}

Task.WhenEach returns an IAsyncEnumerable<Task<T>> that yields tasks in completion order. This can improve perceived performance and reduce memory usage because you don't need to store all results before processing.

In production, use Task.WhenAll when you need all results before proceeding (e.g., aggregating data). Use Task.WhenEach when you can start processing results immediately, such as updating a UI progressively or streaming data. Note that Task.WhenEach is available only in .NET 9 and later.

using System;
using System.Collections.Generic;
using System.Linq;
using System.Net.Http;
using System.Threading.Tasks;

public class ParallelAsyncDemo
{
    private static readonly HttpClient _httpClient = new();

    public async Task RunAsync()
    {
        var urls = new[] { "https://api.example.com/1", "https://api.example.com/2" };
        var tasks = urls.Select(url => _httpClient.GetStringAsync(url)).ToList();

        // Process as each completes (.NET 9)
        await foreach (var task in Task.WhenEach(tasks))
        {
            var result = await task;
            Console.WriteLine(result);
        }
    }

    private async Task<string> FetchDataAsync(int id)
    {
        await Task.Delay(100 * id); // Simulate varying latency
        return $"Data {id}";
    }
}