The concept of "ShardRole retry loop" (i.e. execute something and if it throws due to this shard having stale/unknown sharding metadata then refresh and try again) is present in several places throughout the code. Examples of this are service_entry_point, resharding_data_copy_util or ttl.cpp (this last one may be a stretch since the refresh is asynchronous and there's not retry at this level).

This ticket is to consider making this pattern a common utility that can be reused wherever needed.