The following OOM looks dubious to me:

[GC [PSYoungGen: 234388K->101997K(231616K)] 920213K->793744K(930688K), 0.7431054 secs]
[GC [PSYoungGen: 218534K->96940K(233024K)] 910281K->795560K(932096K), 0.2210026 secs]
[Full GC [PSYoungGen: 96940K->15663K(233024K)] [PSOldGen: 698620K->699071K(699072K)] [PSPermGen: 12562K->12562K(16384K)] 795560K->714735K(932096K), 41.3083684 secs]
java.lang.OutOfMemoryError: Java heap space
[Full GC [PSYoungGen: 116544K->25132K(233024K)] [PSOldGen: 699071K->699071K(699
072K)] [PSPermGen: 12562K->12562K(16384K)] 815615K->724204K(932096K), 37.8715795 secs]
java.lang.OutOfMemoryError: Java heap space
Oct 16, 2003 3:52:20 AM org.apache.tomcat.util.threads.ThreadPool$ControlRunnable run
SEVERE: Caught exception executing org.apache.tomcat.util.net.TcpWorkerThread@6
e24617a, terminating thread
java.lang.OutOfMemoryError: Java heap space
[Full GC [PSYoungGen: 116544K->0K(233024K)] [PSOldGen: 699071K->523327K(699072K)] [PSPermGen: 12563K->12563K(16384K)] 815615K->523327K(932096K), 31.8070514 secs]
[GC [PSYoungGen: 116544K->9954K(233024K)] 639871K->533282K(932096K), 0.3380790 secs]


.. for two reasons:
(a) the heap size when the OOM is thrown is less than the
    max of 1G (it's just 932 M).
(b) Of that 932 M, there's still close to 200 M free
    (i am looking at the verbose gc output just prior to the
     OOM error).

So the OOM seems to be unwarranted.

The above verbose log excerpt is from the following Tomcat
run:

machine: j2se-bigapps.west (contact June for the root passwd)
logs under: Tomcatrun.13536.-d64

Here's further information from June regarding where this bug has been
seen:

  I only saw Tomcat OOM on big machine ( j2se-bigapp.west, 20 cpu,
  81GB memory ) with Parallel GC, 64 bits VM. The machine is Sparc solaris 9.

The process is still available on said machine but is
probably useless (one could potentially kill the client
and let the harness loop again). I will let the engineer
who debugs this problem do those experiments.