I had a P4 which could run prime for about 14 hours, perfect temps, but would always fail after about 15 hours... Same test failed every time. Dropped the FSB by 1mhz, and it would sail through it.
There is virtually 'no' maximum time that you can quote for guaranteed stability. The longer you can test the better.
Ok some might say, who cares 14 hours stability thats great... Thing is there were some other applications which went unstable, and all problems were gone when the system was clocked down by 1 mhz.
Incase anyone wonders, it was a 2.4ghz P4 which clocked at 3.2Ghz with 100% stability, FSB 266 totally stable, FSB 267 prime failed after 15 hours, and occasional "unexplained" crashes. ( I suspect it was a FSB wall, as no amount of tweaking, voltage increases or messing around with memory multi's ever got the system past 266 stable)
So no, even 6 hours prime testing isnt really enough. Yes you might get lucky, on the other hand you might end up plagued by random crashes, which you will no doubt forever blame on microsofts shoddy programming.
In reality, a system which is 100% stable will very rarely crash.. Windows has been given a very bad reputation for crashing, which is largely unwarrented, and is actually a side effect of some very cheaply built computers, which are actually not 100% stable.
I've had 'Pre built' computers from big name suppliers which couldnt run prime95 for 24 hours without an error.