IIS App Pool Shows “Running” but Site Becomes Unresponsive Until Manual Recycle

Hello Captain56022,

Thanks for sharing the context. Based on the settings you shared and what I saw when researching the internet, this could be what’s happening:

Your application pool seems to be entering Rapid-Fail Protection mode because the worker process (w3wp.exe) may be crashing repeatedly. The default for IIS is 5 failures in 5 minutes. 

Even if the pool still shows “Running” in the GUI, once those failures trigger the protection IIS will stop serving requests, so the site becomes effectively unreachable. 

If I were you, here’s what I’d try to check first (just to confirm or rule out this cause):

• Look into Windows Event Viewer (System and Application logs) around the time the outage happened, check for “worker process terminated unexpectedly,” “fatal communication error,” or “application pool … disabled due to a series of failures.” That’ll show whether Rapid-Fail Protection was triggered. 
• If you find such failures, you could temporarily disable Rapid-Fail Protection (for testing only) to see if the site stays up, but this is only a diagnostic step, not a fix. 
• Ultimately you need to identify what’s causing the crashes (bad code, config, dependencies, module load issues, identity problems, etc.) and fix that root problem. 

Hope this helps.

If the site still hangs with Rapid-Fail Protection disabled, the fastest way to find the cause is to capture a memory dump while it's hung, before you recycle, and look at what every thread is waiting on. A recycle throws that evidence away.

1. Get dotnet-dump

On a machine with the .NET SDK: dotnet tool install --global dotnet-dump. On a server without the SDK, download the single-file executable from https://aka.ms/dotnet-dump/win-x64.

2. Capture a dump while the site is unresponsive

Run an elevated command prompt on the server:

dotnet-dump ps
dotnet-dump collect -p <PID> -o C:\dumps\hang.dmp

Pick the right process: with in-process hosting (the default for ASP.NET Core on IIS) the app runs inside w3wp.exe, and with out-of-process hosting it runs in dotnet.exe or your app's .exe. IIS Manager → Worker Processes shows the PID for each application pool.

3. Look at what the threads are waiting on

dotnet-dump analyze C:\dumps\hang.dmp
> parallelstacks
> syncblk
> threadpool

• parallelstacks groups threads with the same call stack, so you'll see something like "40 threads all waiting in YourService.GetOrders".

syncblk shows which thread owns a lock other threads are waiting for.

threadpool shows whether work is piling up in the queue (thread pool starvation).

You can also open the .dmp file in Visual Studio and use the Parallel Stacks window.

What to look for with this "only when quiet" pattern

A background job (a timer or IHostedService) that takes a lock and then blocks on a call without a timeout.

Blocking on async code (.Result, .Wait()) that ties up thread pool threads.

An outbound call (database, HTTP, Redis) whose connection was dropped while idle, for example by a firewall, and that has no timeout, so the first request after a quiet period waits forever.

I reproduced the first pattern in a .NET 9 app: a timer took a lock and then blocked, and every request needing that lock hung while other endpoints still responded. In the dump, syncblk showed the one thread owning the lock, and parallelstacks showed that thread inside the timer callback with four request threads waiting on the same lock. That points straight at the line to fix.

If you capture a dump, feel free to post the parallelstacks output here and I'll help read it.