Quote Originally Posted by dbasan
Increase the db_cache_size to reduce your buffer waits and you can see some improvement.
Next thing you know you'll be charging $300 and hour and have your own remote-dba service.

Quote Originally Posted by slimdave
this seems like a clear case for using statspack to determine the major waits on the system after startup and then after three hours.
Well, sort of. I'd take one while things are "fast", one when things started going "a little slow" and then every 10 minutes until things were "slow".

I don't see how stale statistics could be relevant
I agree, unlikely.