Thor queue has two Thors in it and if job ran on second cluster then Thor Slave log download does not work. It seems
Owned<IGroup> nodeGroup = queryNamedGroupStore().lookup(groupName);
fails. It may be that the other Thor works only because its name is an exact substring of the queue name.
An example is:
queue name is: thor20_167_dev_dev
thor clusters for this queue are: thor_20_167 and thor20_167_dev02
If job ran on dev02 then Thor Slave download fails.
This is from: http://alpha_dev_thor_esp.risk.regn.net:8010 for various WUs.