When we interact with the storage layer and cause it to bring a new page into the WiredTiger cache there is a change that under cache pressure we need to retry backoff if there isn't enough cache to fit the new page. the classic engine has code to do this here https://github.com/mongodb/mongo/blame/528d3eebf23c58beeb03fe6a500d2e6a3851fd58/src/mongo/db/query/plan_executor_impl.h#L95. After looking at sbe::ScanStage and similar we only check if yield is needed and don't expect this exception to occur. This results in potentially returning temporarily unavailable to users which may be unexpected