It's only good for really short strings (< 20 bytes), but our strings are not that short:
- node IDs are not that short:
quickwit-indexer-prod-us-east-1a
- index UIDs are not that short:
logline.c06d8f1aee.2024-02-01.10:01HNGZE70MS3YBV8FVSMVTA910
- queue IDs are not that short:
<index_uid>/<source_id>/<shard_uid>
Let's revert to the default HashMap/Set.