One possibility:
safe_strings was committed fully on 31/07/2009 (there were a few days after the 5492 commit where we fixed bugs).
Depending on how it was built, safe_strings will usually just crash without a warning in Release builds (it's turns out to be quite difficult to catch the assertions raised from the MS CRT - I only had a brief look at how to do that)
In debug builds you get all kinds of spew about over run buffers if that's the cause.
There's also a fix for a CTD which can occur from a call from the PDB_DEBUGGING code which was committed 2/09/2009 in 5557 for clean_filename which could be buggy (hey, no more mangled symbols!)
Beyond those two, we probably need enough info to actually get this to occur locally so that we can track down what exactly is going wrong.