Investigate: can garbage accumulate due to deletion process?

[a-type]

Deleting a doc issues delete ops for all sub-entities the client knows about, but couldn't a peer be creating other sub-entities concurrently which aren't included?

Does this all come out in the wash of history compaction, or do those operations get stuck?

My gut says it doesn't. Unconfirmed creates won't have baselines, so no worries there. The ops will eventually be confirmed and applied, forming new baselines. Those baselines are 'headless' and won't be cleaned up by any process I can think of... And will never get delete ops of their own.