Replication ignoring column which is missing on slave

Replication appears to silently fail on a table which exists on the slave that is missing a column that is on the master. The master is part of a 3-node Galera setup. The column was created during an rolling-schema update which makes me wonder if that's the reason the column never made it to the slave. Row based replication is of course being used here.

The relay log on the slave includes the column when doing a 'mysqlbinlog -v -v'. The column is the last column on the table.

The table is getting updated on the slave, minus the column. So it looks like it's just silently ignoring the missing column from the replication stream but writing the rest of the data? Slave_skip_errors is empty and slave_sql_verify_checksum is enabled.