I recently encountered a bug on my production postgres which results in killing replication slot. It happens when there is an UPDATE statement called on a record having text column with a very long value (about 1mb). The problem can be easily reproduced by this query:
Here I am using decoderbufs decoder https://github.com/krzychk/decoderbufs but it doesn't seems to be related with this plugin as it also happen on the postgres test decoder.