Premature wakeup of time.sleep()

[Erich Schreiber]

Thank you every body for your comments.

Especially Jeff Epler for your hint about NTP. You're right. I see a
(constant but somewhat huge)time drift of about 1 ms/min which I can
correct for.

Thank you Steve Horsley for the clarification of the interrupts that
would end the delay. I didn't think of something as trivial as ^C.

And thank you Nick Craig-Wood for the shell script to get hands on the
HZ value. I had to modify it a little as I do not have
/proc/interrupts, but I've got the idea. I measure about 100 and so I
have a 10 ms granularity in my VPS as you predicted.

Erich