Open Mirroring CSV column types not converting?

This could be a bug or an idea if this is not cattered for:

When uploading CSV files to Microsoft Fabric's Open Mirroring landing zone with a correctly configured _metadata.json (specifying types like datetime2 and decimal(18,2)), why are columns consistently being created as int or varchar in the mirrored database, even when the source CSV data strictly conforms to the declared types? Are there known limitations with type inference for delimited text in Open Mirroring beyond _metadata.json specifications?

Are there specific, unstated requirements or known limitations for type inference and conversion from delimited text files in Fabric's Open Mirroring that go beyond the _metadata.json specification, or are there additional properties we should be using within _metadata.json to force these specific non-string/non-integer data types?