ok fair enough, however a caveat is that it's not necessarily end of function as I was able to expand the function by a few more lines and the callstack stayed the same, however I'm guessing the interpreter was smart enough to realize the calls below the try/except had no references.  

it would be nice if there was a virtual "socket.__del__` or something at the end of the stack.  Basically if it could plug into the extension callbacks.  That way we could have a little more visibility.

closing, thanks guys. This fixes the issue in botocore, on to the next related leak found via aiobotocore in aiohttp, where it now has no tracemalloc entries, so I'm guessing a leak via the ssl module to openssl :(

thanks again guys for the help, I really appreciate it, I hope in the future, my some mechanism, scenarios like these will be a lot easier to decipher.