As Postgres 9.6 reaches its EOL in November, the need for keeping around support for wal2json will be largely reduced. For Postgres 10 onwards, pgoutput is the recommended configuration on any managed Postgres offerings, with decoderbufs still being a fallback option for on-prem deployments with the ability to add custom logical decoding plug-ins to the database. On RDS, remaining 9.6 instances will be upgraded to PG 12 early next year.

In that light, let's consider to remove wal2json support in Debezium 1.8 (released in Q4). It has proven to be a constant source of subtle errors, complexities in the connector code and increased maintenance cost. Removing it will substantially reduce the maintenance burden for the Debezium Postgres connector going forward.
This should be done hand-in-hand with DBZ-703.