Odd problem with alpha9

[Dana Canfield]

Here's an interesting problem, that I would be most grateful for some help
with...

We recently upgraded from a homebrew P233 box to a VAResearch PII 400 box,
both running RHL 5.1 with Intel EEPro 10/100 cards.  When I moved
to the 400 Mhz box, I installed the 2.0.0-alpha9 RPM that
I had been using from the old box to the new one, and I then copied the
MACHINE.SID file and the /etc/smbpasswd file over.  What happens now is
that the Samba PDC will run for about an hour or so, and then it will seem
to hang.  Clients will no longer connect to the box (connections timeout),
and existing connections sometimes talk about semaphores expiring, but
just generally quit working. The smbd continues to fork, though (as
though it is receiving requests), until it blows the machine to oblivion.

If samba is stopped and started (before crashing the machine), that isn't
sufficient to make it work again.  Instead you have to stop and re-start
the network, then restart samba.  Then the whole process repeats.  I tried
upgrading to alpha12, but neither box will operate properly with that
version, and 13 doesn't compile cleanly.

The new machine runs fine for everything else (including netatalk, which I
figure is as strange as anything for network protocols).

Does anyone have any idea what might be causing this?  I'm particularly
confused because I'm running the same OS, same samba binary, and same
ethernet hardware on both machines.

For now, I'm just using NFS and letting the old box do NT_DOM stuff, but
I'd appreciate any thoughts on why this could be happening.

Thanks!

Dana