We have a process that runs once an hour to read the .dat file in csv format then a node script using the pg package version "8.8.0" to create the json objects and insert the data records as jsonb data.
None of the upload process changed during the underlying database upgrade.
Basic debugging requires the existence of a self-contained reproducer. In this case ideally one that only uses psql and some static (already processed) data files, and that is known to produce the observed behaviors on non-RDS PostgreSQL.