Clock adjustment should be detected and tolerated

What steps will reproduce the problem?
1. With the cluster running, make a major adjustment to the system clock 

What is the expected output? What do you see instead?

Expected: It should tolerate a clock change; some timing information may be
lost but it should continue to function correctly.

The master node should not expire any other nodes even if the clock jumps
forwards. All nodes should continue to run Tick() at approx 1 second
intervals regardless of clock jumping.

We can easily check the timestamp of our last tick, and if it was a long
time ago, or in the future, assume that a clock adjustment has happened and
tolerate it. All nodes need to have their last heard time reset (as it will
now be meaningless).

Original issue reported on code.google.com by mar...@gmail.com on 1 Jun 2008 at 1:05

deriamis / fluffy-linux-cluster

Clock adjustment should be detected and tolerated #5