This can lead to the node that went through rollback reporting a different capped collection maxSize than nodes in the cluster that do not rollback.

How we get there: 

• At some point after loading the catalog during rollback state
• validate the catalog entry
• validate and parse for storage
• check and adjust capped size
• round up to multiple of 256 bytes

This is affected by refreshing the in-memory FCV value to reflect the post rollback state.