I have a theory that the reason why we see some test flakiness is because we are running tests in parallel on already starved GitHub actions VMs. This issue is about testing that theory and removing flakiness.
Some ideas:
Try not running tests in parallel. go test -p ./... instead.
This will significantly slow down our test times. We could offset this by caching test results and only testing things that have changed.
I have a theory that the reason why we see some test flakiness is because we are running tests in parallel on already starved GitHub actions VMs. This issue is about testing that theory and removing flakiness.
Some ideas:
go test -p ./...
instead.