I would imagine that the intended use case for this source would be here, when calling into performUpdates in order to compress an existing uncompressed bucket.

However, we use kTimeseriesInsert here, and it does not appear that there are any places where we set the source to be kTimeseriesBucketCompression. Despite this, there are places where we have specific error handling for the case where the source kTimeseriesBucketCompression: here and here