chore(qe): update mobc to include importcjj/mobc#86

[miguelff]

Fixes prisma/prisma#21221

[miguelff]

Created as an integration branch, in case some of the tests in prisma/prisma depended on the previous (wrong) behavior described in importcjj/mobc#82

[codspeed-hq]

CodSpeed Performance Report

Merging #4379 will not alter performance

_{Comparing integration/bump-mobc (411821c) with main (e0f0b06)}

Summary

✅ 11 untouched benchmarks

[miguelff]

According to prisma/prisma#21591 there aren't failing tests after upgrading mobc.

[Jolg42]

@miguelff it would be interesting to review tests, we should have had one catching this change ideally, right?

[miguelff]

I don't want to encode in the tests specific values for queries, because depending on pool sizes, those will behave differently. What I observed is that before (main) reported idle = pool size, while now, it reports 1. Which is the expected, correct behavior.

This can sometimes be reported as 1 or 0, and lead to flakey tests, tho' as metrics are reported asynchronously.

main.txt
branch.txt

[Jolg42]

Right!

My thinking is that we could expect 0 or 1 for prisma_pool_connections_idle, I think that's better than nothing. Though not ideal.