Code review comment for lp://staging/~rogpeppe/gozk/safe-close

This essentially LGTM; I don't think that it's a serious problem that
the tests can theoretically fail under "enough" load. So it comes down
to a question of practical reliability: for the sake of argument, what
if we shoot for an observed failure rate of <1% on your local machine
under "normal" load, and bump up the sleeps as needed if they turn out
not to be good enough for clean testing in practice (say, on ARM ;))?

https://codereview.appspot.com/5699093/