Is there any way to be more specific about the target schema/column types? E.g. decimal types. Depending on the particular set of records captured in a CDC (log-based) run, I'm getting incorrect and inconsistent inferred types. e.g. int64 if all the decimals are coincidentally integers, or int32 if everything is NULL. Editing and passing in a catalog.json doesn't seem to help as it's ignored e.g. (["string", "null"] still becomes an int32 if the values are all NULL, and rounded decimals still become an integer). This inconsistency breaks any notion of having a fixed schema that I query the destination from. Wanting to avoid having to push everything to a staging area, patching all the column types, then pushing them to the actual destination.

Have you tried using https://docs.meltano.com/concepts/plugins/#schema-extra? I see you mention

Editing and passing in a catalog.json doesn't seem to help as it's ignored

. I think that's a major gap in the tap.