find stuck in tstile

[steve%prd.co.uk@localhost (Steve Blinkhorn)]

I thought this was an issue that had gone away, but...

It concerns /etc/dail and /etc/weekly and the complex calls of find(1)
they generate.   These invocations get stuck in tstile, and associated
shells, tees and sendmails also accumulate, so that a server that
might normally be runnjing at about 75 'active' prcoesses ends up with
several hundreed.  I can get rid of the shells, tees, sendmails etc,
but not the find processes.

This used to happen on my old i386 servers running 7.x, but now it's
happening on my shiny year-old amd64 servers running 8.0.  It only
seems to happen on my colo machines, not on the clutch of machines in
the office.   The colo machines have a lot more disk space than I use.
It's very unlikely to be a fault in an individual machine - I have two
identical servers next to each other in the rack with the same problem
at the same magnitude at th same time.  One primarily serves
customers, the other is primarily a backup machine.   We had the same
issue with i386 servers running 7.0, and as I now realise it was also
/etc/daily and /etc/weekly that caused the issue there.

If I try to run df(1) it gets stuck in tstile too and I can't kill it.

Apart from periodic rebooting, can anyone suggest an answer?  I used
to run servers for years between reboots.

-- 
Steve Blinkhorn <steve%prd.co.uk@localhost>