I see that derby.log contains many occurrences of this message before the test fails:
------------ BEGIN ERROR MESSAGE -------------
new log file exist and cannot be deleted C:\cludev\jagtmp\autoderbyN_derbyall\suitesAll_1\log\system\wombat\log\log454.dat
------------ END ERROR MESSAGE -------------
The test run had a failure in StreamingColumnTest (an intermittent failure fixed in
DERBY-4531 but not backported) before the problems in TriggersTest, so perhaps a file handle was left open and caused problems when later trying to write to that file? It's a bit strange, though, that so many other tests ran in between without any issues, but it may of course be that none of those tests produced enough log files to hit that file.
Hmmm... So it seems this log454.dat file that we cannot delete, prevents us from switching from log453.dat, and we just continue appending log records to file 453 until it exceeds the maximum size. Would it be a more appropriate response to fail immediately in the case where we cannot switch log files, and tell the user to force the removal of the file that's blocking us?