Near-duplicate schema created from WAL vs. snapshot numeric type with column.propagate.source.type

I have noticed an issue of multiple schemas in schema registry being created for a particular kind of table: one with a numeric (unbounded) type, using column.propagate.source.type.

The following picture shows an example of the schema diff in the Confluent UI.  As you can see, the only difference is that the original schema contains scale, whereas the new one does not.

Steps to reproduce:

1. Create a table with a primary key and numeric type.  Populate with 10 rows with some NULL values for the numeric fields, and some with value like 1.5.
2. Add this table to Debezium, including a snapshot
3. Insert a new row, also with NULL for the numeric field
4. Observe a second schema being created in schema registry, without scale.

I am not sure if this only happens with respect to NULL values.  I'm also not sure if this is by design or not.

But I would think that the schema in schema registry for Debezium should never change unless there is an actual data type change in Postgres?

Please let me know if I can provide more information.