One of my customers have three MSCS (Active-Passive) 2-node clusters under Windows 2003 SP2 Standard Ed. The nodes run one instance each for SQL2000 (SP4) and SQL2005 (SP3). They run DBCC CHECKDB on all databases, and also backup the logs once an hour. Periodically throughout the day, like every 30 min, the "I/O takes longer than 15 sec" message appears in the SQL Log. The servers are HP Proliant (BL25p), each have 8GB of memory, storage is EMC. The servers and storage checked out OK, Did a Storage check, couldn't find any latency anywhere. Besides the I/O error, once a month or so, the SQL instance itself crash (generally no meaningful errors), but would restart itself. But once in 2-3 months, with on SQL Instance failure, the node would failover.
This behavior is similar among all 3 clusters. The Windows EventLogs, SQL ErrorLogs show nothing tangible.