We had a job create some “duplicate” events, and our domains aren’t doing versioning, nor doing things idempotently, so we ended up with two events with different IDs, same body, in multiple streams.
The plan was to soft delete those streams and write new, filtered versions of those streams.
We then reindex the store into Elasticsearch using $all.
The streams themselves delete as expected, returning a 404 with an HTTP call (but isstreamdeleted=false in TCP client metadata). However while slicing through $all, it seems the deleted events still surface there.