Adam Williamson wrote:
The good is not the enemy of the perfect. A 90% chance of noticing a
problem is still better than a 10% chance, even if it's not 100%.
That doesn't help when this takes a week and you break a dozen machines
while waiting for the "don't destroy hardware!" fix (and that's just if
the
hardware is really infrequent; you could end up with a lot more broken
machines in a week even with frequencies which may still elude testing).
Kevin Kofler