Jason Madden

I should have added: I find it very interesting how the line that was reported to leak in ssl.py *changed* when removing C accelerators. That doesn't make much sense either.

Thanks for your patience and assistance debugging this. I'm suspicious of bugs in the underlying `greenlet` library. I am certain I found one leak, but it only affects free-threaded CPython,...

There's not enough information here to suggest what the problem is or what a fix would be. Can you supply a [short, self-contained example](http://sscce.org) that reproduces the issue? Unless you...

Based on that backtrace, I theorize that you're filling up gevent's thread pool because you're spawning way too many greenlets, all of which are, essentially, waiting on each other. If...

> We are actually passing a timeout to the requests.get call Somewhat immaterial, as that doesn't provide a timeout for the *entire* operation, only the part where you're reading data...

Just to clarify, your sample code isn't monkey-patching, but the provided stack trace appears to be from a monkey-patched system. I assume you are actually monkey-patching? And hopefully not setting...

The concerning thing is that all of the worker threads are _dead_: ``` ; not running : Parent: ``` Something is killing them that isn't getting caught and handled. Figuring...