Title says it all really. This bug was detected by Ferdinand over the summer, and I've moved discussion of it onto the bug tracker. The input file which causes the crash is attached. It crashes when placed in the 4h128 queue... haven't tried anything else yet.
We had some thoughts it was to do with diagnostics, but it is not because it is reproduced when loop_diagnostics is commented out.
The error messages are varied, sig 9, sig 11 both seen.
The smaller the problem, the longer it runs before crashing which does indicate some kind of memory leak.