Q about named not responding the TCP queries

[O'Neil,Kevin]

We had a problem where named 9.3.5-P1 running on a Solaris8 box stopped responding to TCP queries while still responding to UDP queries.  There was still a listener on TCP port 53.  It took a stop/start of named to correct the issue.  

At the time that this event occurred we saw this in the messages file:
Nov 20 13:35:31 ... named[29028]: [ID 873579 daemon.error] internal_accept: fcntl() failed: Too many open files

Has anyone encountered this situation before?  What was the cause?  What was there a fix/workaround other than updating BIND?  Is this an issue with the underlying OS?  Just increase ulimit?

Searching on the web didn't get results dealing with this particular situation. 

Thanks.