DHCP dies after sending NAK

[Norman Elton]

I have a reproducible bug, where the ISC DHCP server (v. 4.1.0p1)
stops processing packets. I've been able to reproduce the problem
under the following circumstances...

The server is authoritative for a subnet, and accepts requests sent
from a DHCP relay. Clients broadcast requests, the relay forwards them
along, everyone is happy. At some point, a client decides to unicast a
renewal request. That is, a DHCPREQUEST comes directly to the server
without the "giaddr" attribute set.

Here's the output I'm getting...

DHCPREQUEST for 192.168.38.33 (128.239.10.33) from 00:11:22:33:44:04
via nge0: wrong network.
DHCPNAK on 192.168.38.33 to 00:11:22:33:44:04 via nge0

That's the last line my logs show. Further packets seem to be ignored
until I manually restart the server.

I found a similar problem in an older post
(https://lists.isc.org/pipermail/dhcp-users/2009-March/008475.html),
but am not sure if the problem was resolved. My server is also running
on a Solaris box, is this something that is specific to their network
implementation? I might try moving to another OS.

Any other ideas?

Thanks,

Norman