Joseph Bylund

> > How about adding a test to the finder which demonstrates that calling `find_best_candidate` after calling `find_all_candidates` results in a cache hit? > > How difficult it would be...

Looks like this is the failure: ``` 2021-10-21T19:03:24.7123305Z nox > Running session docs 2021-10-21T19:03:24.7469283Z nox > Creating virtual environment (virtualenv) using python in .nox/docs 2021-10-21T19:03:25.8502672Z nox > python -m pip...

Looks like the multiprocessing wrappers were dropped at the same time as support for 3.6. Does that mean we can use `multiprocessing.ThreadPool` directly now? I won't be able to respond...

@uranusjr do you need anything else on this one?

> LGTM. The only comment I'd make is would it be worth pointing out explicitly in a comment that this relies on the fact that `find_all_candidates` is cached using `functools.lru_cache`?...

> We have an [in-progress PR](https://github.com/urllib3/urllib3/pull/2661) which defers the closing of connections at the HTTPConnectionPool level until the connection pool is no longer referenced instead of during PoolManager eviction which...

@bluenote10 thanks for the clear reproduction instructions, makes it so much easier to look at something like this. I took a profile on my machine and found that it was...

> * I can `cp -ar` the entire pip cache directory in ~6 seconds (for 3.4 GB). > > * The pip install takes ~92 seconds, which feels > 10...

> Please share the requirements.txt that you're experimenting with to reproduce this. Should be in the first comment in this issue.

I haven't profiled what you describe yet, but it sounds like a separate issue with a similar symptom. If others agree I'd propose re-naming this issue to be about parallelizing...