S2S get_missing_events: does the walk terminate at `earliest_events`?

Nobody has claimed this yet.

**Link to problem area**: https://spec.matrix.org/v1.6/server-server-api/#post_matrixfederationv1get_missing_eventsroomid

**Issue**
Says:

> This is done by doing a breadth-first walk of the prev_events for the latest_events, ignoring any events in earliest_events and stopping at the limit.

And for `earliest_events`:

> Required: The latest event IDs that the sender already has. These are skipped when retrieving the previous events of latest_events.

In both cases it's not clear what exactly is meant with "ignoring" and "skipping" - does this simply mean that those events are elided from the response, but the walk continues beyond them? Or does it mean that encountering an event from the `earliest_events` terminates the DAG walk for that specific branch, and so any known event's ancestors are not considered either (unless otherwise encountered)?

Start with the Matrix Server-Server API's POST /_matrix/federation/v1/get_missing_events section linked in the issue. Compare the descriptions of latest_events and earliest_events, then establish the intended walk behavior with the specification maintainers; done means the endpoint text unambiguously defines how those events affect traversal.

Written by the indexing model from the issue text.