I resolved the last issue by deleting the repository and installing the latest update. I deleted the database and had it start fresh.
It worked since the 17th, but now it is back.
I am trying to figure out what is going on.
Right now the query running for 40 minutes is:
FROM @Chunk c
INNER JOIN [data].[Cluster_Keys] d WITH (ROWLOCK) ON c.[Id] = d.[Id]
-- FORCE ORDER: The order above is carefully chosen, and this prevents pathological orders.
-- LOOP JOIN: Profiling shows that this tends to improve performance, as long as the number of rows to be deleted is small
-- compared to the number of rows in the table, by encouraging index seeks.
OPTION (FORCE ORDER, LOOP JOIN, MAXDOP 1)