maurycy

@flavorjones Thank you as well! That's true. I had to manually `byebug`.

@flavorjones While we're at it, what do you think about covering [ssl_timeout](https://www.rubydoc.info/stdlib/net/Net/HTTP#ssl_timeout-instance_method) as well? It's [well supported by net-http-persistent](https://github.com/drbrain/net-http-persistent/blob/75574f2546a08aa2663b06a2e005bcf2ee304f13/lib/net/http/persistent.rb#L961). I can handle this. [Net::HTTP.keep_alive_timeout](https://www.rubydoc.info/stdlib/net/Net/HTTP#keep_alive_timeout-instance_method) is [shadowed by Net::HTTP::Persistent.idle_timeout](https://github.com/drbrain/net-http-persistent/blob/75574f2546a08aa2663b06a2e005bcf2ee304f13/lib/net/http/persistent.rb#L600), so not...

A simpler way to reproduce: ```python import pyperf def fake_bm(): pass runner = pyperf.Runner() benchmarks = [] benchmarks.append(runner.bench_func("first_run", fake_bm)) benchmarks.append(runner.bench_func("second_run", fake_bm)) suite = pyperf.BenchmarkSuite(benchmarks) ```

> Since pyperformance users include not only CPython but also alternative implementations, I don’t think relying on uv would be a good approach for us to take. cc @hugovk `uv`...

@mdboom Does the optimal number of runs / processes might vary? For example: less runs needed on Linux than on Windows, or different variance on PyPy? If you still have...