3 ms·
Also add that 'flake tests' were a sore point until it became clear which ones were actually the system failing and which tests didn't reflect actual usage. By
by karmakaze 8y ago
Also add that 'flake tests' were a sore point until it became clear which ones were actually the system failing and which tests didn't reflect actual usage. By keeping stats on test faulures it was clear which were causing distraction time and fix them or system cause first. After that we got stability much more quickly than always treading seemingly random errors.