Last week I encountered a strange phenomenon in a RAC cluster I'm concerned with: Some statements on node2 showed a drastically bad performance when compared to the same statements running on node2 (the AWR reports showed a ratio of 50:1 for the seconds per execution of the same statement). After node2 was restarted, performance of the two nodes was roughly the same again. From the AWRs I can see that node2 had high wait times for db file sequential read.
My expectation was that performance of one node would degrade again over time, but until now, everything looks well. I have not been able so far to identify a singular condition on node2 that might have caused the deviation: No additional jobs running, no hanging backups, nothing special in the alert logs.

Where should I turn next for my analysis of this incident? Any hint appreciated!