coverage: 72.345% (-0.02%) from 72.369%
when pulling 2460100 on alumni:fix/sqlite-flaky-test
into 79960e1 on typeorm:master.

i think i can help you with creating a test that can reproduce the error.

The issue I referenced might not be affecting this test.

I think I know the error for the flaky tests (it's the one in the PR description and I can reproduce it easily), it's caused by how we work with savepoints in the test - we can't have parallel savepoints, we can only parallelize the top transaction, not the lower level ones like the test was doing before. And this should be true in general in SQL, regardless of the dialect.

I actually don't know how the test worked for other db drivers - this is giving me a bit of doubt, so some confirmation that what I stated above is true or false would be helpful.

I didn't have time to look further, that's why I didn't update this anymore. The last issue I had after the changes in the test is that I couldn't parallelize those transactions (they go into deadlock for some reason). I'll come back to it later this week, but feel free to suggest anything or even take over if you'd like.

Declining this because I'm reverting the PR that is causing our issues in #11264 and wrote an additional test there.