Problem resolved!

Databricks Solutions Architect Casey Orr suggested that having 2 different versions of the ddl audit table and trigger in place was the issue - and he was right.

As we have been using Lakeflow Connect for more than six months we have seen a number of versions of the change tracking scripts - most recently with version 1.5 - and it looks like there was a misunderstanding on our part as to which scripts are still current, meaning that one of the legacy scripts - that creates the 1.2 version of the ddl audit table and trigger - being run after the 1.5 version.

This seemed to confuse the LC service (not sure why, as everything is versioned now, but still) and simply dropping the older 1.2 version of audit table and trigger got everything running again.