Long running test spawn too many threads

[sjmonson]

The SyncManager process spawns a huge number of threads over the lifetime of the benchmark. On pid constrained systems (such as OpenShift) we can run out of threads during longer tests or with shorter requests.

[sjmonson]

This issue still occurs after replacing the SyncManager with shared queues.

[sjmonson]

After a bunch of testing I've found that the issue only occurs with Python 3.12.