GC timeouts on stress runs on excalibur

Barrier timeouts have become more frequent in recent weeks on the gcstress runs on excalibur. I've looked at half a dozen of the failures, and they all have the same diagnostic program point information (typical example appended from: http://jikesrvm.anu.edu.au/cattrack/results/excalibur.watson.ibm.com/stress-ppc32/4871/gcstress/Opt_1/dacapo/jython

GC Warning: Barrier wait has reached 26.64994987 seconds. Called from GC Warning: Barrier wait has reached 26.64994991 seconds. Called from 1004. myOrder = 0.00000000 count is 1004. myOrder = 3.00000000 count is 4 waiting for 2
1 waiting for 2
GC Warning: Barrier wait has reached 53.30152090 seconds. Called from GC Warning: Barrier wait has reached 53.30163692 seconds. Called from 1004. myOrder = 3.00000000 count is 4 waiting for 2
1004. myOrder = 0.00000000 count is 1 waiting for 2
Died in GC:
GC Error: Barrier Timeout