HandleResponse called on URL which is already erased

[GoogleCodeExporter]

There is a known bug where, under load, we end up calling HandleResponse on a 
URL that is already erased.  This currently causes a fatal check failure in the 
Apache process, with a message pointing to this bug.  Without the check 
failure, the apache process will livelock and eat all available CPU; we've 
chosen to crash and recover because it is less bad than livelock in practice.

We're interested in obtaining distinct stack traces produced by this failure in 
order to better understand the circumstances under which it occurs.  Please 
report the version of mod_pagespeed you were running when you encountered the 
problem.

Original issue reported on code.google.com by jmaes...@google.com on 25 Feb 2011 at 7:08