I have a weird issue.
I am running github actions on self hosted runners in google cloud so that i can spin up an emulator.
We have sometimes a flaky tests which fails.
Recently we saw that when we want to rerun the job, we get lots of time out of memory exceptions in the instrumentation tests of the emulator but when we close the pr and reopen it and retrigger the job the instrumentation tests pass well.
What i don’t understand is that both start a clean job so why do i get these discrepancies?