In retrospect, while doing this on-wiki versus in Gerrit made it easier for you to add implementation notes, it makes it harder to comment on things. Flow makes it even more of a pain.
slot_inherited seems more natural to me. A content row could theoretically be reused for later revisions. For example, if someone makes and edit and that edit is reverted using the rollback feature, we're probably going to take advantage of that trivial case to reuse the old content row. With content_origin, though, we'd either be prevented from doing that because we'd have to update the origin or the existing origin would cause the slot-history thing to skip showing the revert.
slot_inherited would also be more efficient for your proposed use case, since it won't require joining the content table to determine whether a revision can be omitted.