Bug Report: execution time increased by ~33% with 3.12 image

[alexei]

Describe the bug

We upgraded to 3.12 yesterday and found that the execution time increased from ~30m to ~40m. I re-ran on 3.11 in a different branch earlier and found it took ~33m, in line with earlier measurements.

For comparison, I ran the tests locally and got ~27m on 3.11, and ~25m on 3.12.

To Reproduce

Not sure, as this is a private project.

Expected behavior

I would expect to see similar if not better results on 3.12.

Screenshots

This is how master looks like for the past 7 days:

Additional context

I'm curious whether this is something that others have experienced, and what are some steps we could take to debug this etc.

[alexei]

One of my team mates suggested this might have to do with https://github.com/nedbat/coveragepy/issues/1665 I did not run coverage locally, I'll try to do it tomorrow to compare.