Hi Vojtech,
I changed my mind a bit.
Yesterday I wrote
> ... the conclusion is that in reality sync was never lost.
and probably that was also true, at least the system log suggested this.
Yesterday I got half a dozen bursts of pairs of "sync lost" during e2fsck.
Today, with 5*HZ instead of HZ/20 I still got a "sync lost" once -
throwing 2 bytes away. So, sync was really lost.
(Again during e2fsck - 2.5.44 crashes when I invoke yast, an ide-scsi
problem, and I tend to forget this.)
This means that I no longer have an opinion about the appropriate
time interval here. Made it 1*HZ for myself for the time being.
Andries