Both re-crunching a failed wu and getting the same wu if you just delete the file/folder are normal.
If a WU eue due to an internal failure, it will frequently upload the error code and restart the simulation, it this fails a second time at the same place with the same code, it will again upload the error code (the server then 'knows' it a bad wu and I resume it gets logged to be anaylesed) but some time it restarts again!
I've crunched a few WU three time all failing at the same point before getting a new one.
This is part of crunching beta units, thankfully most member only have it happen a few times a year.
It is possable to reinstally the smp client - therefor new ID string and new wu. BUT that just means someone else will get the dodgey wu 2or3 times. so good for you, bad for the community and science.
If a wu dies due to a computation error or checksum - you usually get a new wu, as it eue due to hardware and your wu is sent out to someone else.
Hope that helps - lost over 4.5 days crunching due to one duff smp wu two weeks ago on the workstation here
I got really miffed the first time it happened. I think i'd only just got a new e6420 at the time, and was really pleased with the boost in production, then it all when pear shaped.
These things happen from time to time, it's one reason why the smp points are so high.