Thank you all for the replies.

I will investigate changes that may be relevant in CUs that are different between the instances.

Regarding the TABLOCK recommendation, thanks, but it doesn't apply here unfortunately. The databases was most definitely corrupt. 1x PK had to be rebuild by a repair_with_data_loss, and then a second run of the same was required to fix a few broken pages. Errors were returned when querying the 'broken' table, so not snapshot-related I'm sure.

I would also suspect the IO subsystem with this in mind, but consider that our SAN infrastructure provides storage for over 1000 databases on over 100 instances, but only this one instance has ever had DB corruption in the 4 years I've worked as a DBA here. It's also only this one database...

The last time it happened, a dump file was produced by SQL, and MS said they could find no reason within it for a corruption. We are now barking up the VMware tree.

I'll update this if I find the answer.

Thanks again all