BIND query queues/timeouts

[Tim Maestas]

We're running bind 9.3.4 (also see the issue described below on 9.4.2 as
well) on a dual CPU SUN V210 running Solaris 9.  Every so often we are
seeing a spike in query traffic to the main IP of the servers that clients
point to for recursive nameservice.  After this, BIND appears to be unable
to answer new queries, or is answering them very slowly.  CPU usage goes
up, but there are still idle cycles available.  During this period, when
BIND is not responding off of it's main interface, it IS responding on the
loopback interface, or on another VIP address on the server not used by
many clients.

Does BIND maintain separate query queues per interface?  Is it possible 
that there is a flood of incoming queries that get queued up and the box 
is slowly slogging its way through them?  The bulk of the queries from 
what we can tell are from zones the server is authoritative for.  Is there 
anything else anyone knows about that would cause BIND to bog down like 
this?

-Tim