Dmitry Vyukov

> If report says CORRUPTED: Y, should I expect report.Corrupted to be true? Correct. Such reports will be thrown away and not reported as bugs. But these reports are not...

We may try to remove t.Parallel() from there, maybe there is some contention...

> In any case, I've tried to reproduce it many times locally and never managed to. Have you tried with `taskset -c 0` or `0-1`? The test VMs should be...

Good point. There are sleeping goroutines in connectionLoop in the dump. The simplest thing we can do is to send empty signal updates before the sleep and check errors.

Generally there are always some fault injections happened in the past. How can we understand a past injection is related to bug or not?... We could report injections only in...

> There's no reliable way to know whether fault injection was involved unless there's reproducer (not the case for all our bugs). Can't we understand that a fault was injected...

Do you mean like crashes in sshd? Do you expect kernel crashes in sshd caused by fault injection in our test processes to be a problem?

This does not look like enough justification to drop fault injection. There is no reliable way to do just anything wrt kernel, we don't even always understand if it buged...