BIND cache corruption?

[Mike Buckley]

Hi,

I'm running a DNS server that's displaying some intermittent weirdness.
Every so often it will timeout when running a nslookup/dig to a
particular DNS server, the query doesn't even leave the server. Looking
at tcpdump I can see it sending *lots* of udp53 packets to the target
DNS server (multiple lines per second) without getting a response
(malformed?). This is independent of manual nslookup/dig queries and
seems to be from previous queries. A stop/start of BIND (9.2.1) cures
the problem completely, but it returns every few weeks and the problem
appears to involve random DNS server targets.

It's baffled me a bit, but I'm starting to think that maybe the cache is
getting corrupt in some way. Does anybody know of anyway of
interrogating DNS cache?

Also this DNS server is pretty busy, getting on for 100 queries a
second. It's running on an 800mhz, 512Mb RAM Unix server (AIX 5.1), is
this well within capabilities? Checking RAM/CPU usage shows nothing out
of the ordinary.

Advice? Help?!

Thanks 

-- 
Mike Buckley