I know you're dealing with AC issues, but I would check around AC. How many snaps are you keeping? Number of snapshots can drive higher load, which can negatively affect AC replication. Are you doing any async replication? Larger PGs can take longer to replicate, which can lead to issues that could affect AC. How many IOPS are you typically pushing? Maybe the array is just that busy servicing front-end IO and other services are suffering. Same goes for IOP size. Larger sizes means more buffering, which can affect latency, which can cascade to AC performance problems.