Definitely there should be some errors in logs, in some cases error_log could be stored in same directory where wordpress is installed (or directory where PHP script is located). Check them please.

Also, check if customer is not hitting any lve limits, for example with:

 lveinfo --period=1h --id=UID

Thanks for the suggestion, Bogdan.

The root error_log had nothing, so I did a find on error_log, just in case the application was calling itself somewhere. Found a few errors, all of them months old.

lveinfo is clean:

 lveinfo --period=1h --id=685
+----+--+----+----+----+---+---+---+---+
|From|To|aCPU|mCPU|lCPU|aEP|mEP|lEP|EPf|
+----+--+----+----+----+---+---+---+---+
+----+--+----+----+----+---+---+---+---+

I suspect it's an application error, will look at scanning the code for calls to header() and see if I can find anything. I also tried looking at memory but when the page serves correctly peak use is only 32M.

I\'ve set his site up on a box without CL installed, and it\'s exhibiting the same behaviour, although less frequently. Either way CL has been eliminated as a source of the issue. On to eliminating the next variable...

Thanks again for the help.