>>Some more thoughts:
>>
>>You may need to get MS to clarify what happens when recycling in this situation. According to the docs the old process should not be signalled for shutdown until the new process is up and running. If that were true and the new process fails to start then I'd expect the old process to continue indefinitely but that doesn't seem to be happening.
>>
>>Also be interesting to know whether, when all four processes exist at recycling time, IIS automatically spins up four replacements - or just one and then creates the additional ones on demand. Given that you get multiple error message then, unless the web site is very busy, it looks as if the former is the case.
>>
>>Things I might try on the AppPool:
>>
>>(a) set IdleTimeout to a lower value - this would increase the chances of having less processes to restart on recycle (if the above is true)
>>(b) increase the 'Startup Time Limit'. (Although the default is 90 seconds which should be ample unless something is wrong?)
>>(c) increase the 'Shutdown Time Limit' (Not directly related but would give the old process more time to clear queued hits)
>
>When one or more worker processes do not respond, those are the new ones. The old ones have been removed before that happened.

That was sort of my point - if the old processes are terminated then it would imply that, as far as IIS knows, the new processes was successfully started.
Might be interesting to log Application_Start() and see if it gets that far ? Have you checked the application log BTW ?

>Based on that information, before I answer a, b and c, would you see adjustments on the previous message?

Do you mean turning off compression? Maybe IIS clears out or recreates the folder on recycle (and does that regardless of whether compression is enabled or not)