Tried disabling storage QoS.
We have been unable to replicate the issue still. We've tried generating 100K+ IOPS of all different block sizes and read/write ratios. The disk queue length & latency is very high, but we still can't reproduce the issue.
We have noticed that all but one of the servers this has happened to (about 10 in total now) have had either an MS database or an Oracle database on them. Unsure if related or coincidence.
We have an MS case open, but it's not getting very far as they can't replicate it either.
Has anyone else had this issue?!