Jose wrote:
There are several ways they can work and most of them do in fact do
the math in one way or another. But something has to sense the revs.

How would weakening magnets affect a tach? If they get so weak they
start missing revs (that are counted individually), I would expect the
tach to be very erratic at that point.

Jose

Remember I'm not saying that that is the problem the OP was having but if
you are counting something that happens 2800 every minute (46/sec) and you
start missing say 10% you'd see a 280 RPM drop. Even with that 10%
randomized somewhat the tach could keep up with the change enough for it to
show up as wild swings and would probably not be any worse than that caused
by vibration.