Intermittent get record api failure

We are seeing intermittent get record request failures even though record exist in CosmosDB. The issue is intermittent, cannot be consistently re-produced. Same record that exists cannot be retrieved initially but starts working after some time.

This also has potential impact on Indexer service. We have integrated workflow - all records that are ingested (via storage serivce) must be indexed and discovered via search service. If we have these intermittent failures than indexer service must be configured to re-try in these scenario, otherwise it will lead to inconsistent state.