Hmm .. maybe wrap the offending call in an eval and do a confess if a bad thing happened ..

Interesting what happens when you follow your own advice.

I wrapped the three possibly offending calls up with an eval and after trying to provoke the error finally got *two* errors:

• Inappropriate ioctl for device
• Stale NFS file handle

with the first error causing the second error. So that tells me that it's an NFS problem, which is fine. The parallel processes are jobs running on a Sun Grid Engine that are all using NFS to come back to my workstation's file system.

That tells me that NFS may need tuning, but it's not the fault of the Cache::Cache family of modules or even my web application.

So, problem solved. Thanks anyway!