The only difference I saw was in the “remarks” for the two tables, with one of them saying "btree comparison function" which is strange.
This would seem to indicate that the catalog pg_description has two rows for this particular table and thus the join to it [1] causes the single pg_class entry to become duplicated.
Per the comment command page each object gets at most one comment so having multiples in the catalog is data corruption. If you confirm that you do indeed have duplicates hopefully issuing create comment on the problematic records clears up the issue.
