I read Rocket Scientists cluster blog and I can synpathize with him.

We updated our Compaq cluster to the latest firmware so we could install SP4 for Win2k.  12 hours after install transient error 823 began to occur.  The application would get the error but by the time we could start a dbcc the error would be gone. We fought this for about 2 days working with micorsoft. They finally suggested turning off the caching in the controller.  We turned it off and have not had an error since.  According to Microsoft  SQL2000 SP3 has issues with some caching controllers. They are not sure if it is stale reads or writes that cause the error. They have a new hotfix with enchanced logging traceflag 818, but since this is a 7x24 production system and all the mew version does is log errors that I can send to compaq, we did not install it.  Funny thing with all the caching disabled we have seen no performance issues, but Compaq has warned us that if we have a hard down the non sql data on the drives may be at risk for corruption.  Our next step is to renable the caching on the non sql drive arrays and see if we have any problems.