Locked index with a changed definition starts silently with the old definition or fails with an unclear RavenDB error #6096

Describe the bug

The documentation recommends a lock on a migrated index. When ServiceControl ships a changed definition for a locked index, the change is not applied until the operator unlocks the index, and ServiceControl does not report that:

• A LockedIgnore index starts silently with the old definition.
• A LockedError index fails start-up with a RavenDB exception that does not name the index.

The operator cannot see that a definition change was skipped, or which index blocks the start. ServiceControl also does not report which indexes RavenDB rebuilds at start-up, or which search engine each index uses.

Steps to reproduce

1. Lock an index in RavenDB Studio with mode LockedIgnore.
2. Start a ServiceControl version that ships a changed definition for that index.
3. ServiceControl starts. The log does not mention the index, and the old definition stays active.

Relevant log output

No response

Additional Information

• Affected versions: 6.20.0 to 6.22.1
• Related: Warn operators before an index definition change triggers a RavenDB index rebuild #6015
• Fixed on master by Report locked indexes and index rebuilds at start-up #6036

Backported to

• 6.22.2: Locked index with a changed definition starts silently with the old definition or fails with an unclear RavenDB error #6092