Perhaps ar_parent_id could be populated for rows with ar_page_id set, such a script could check both the revision and archive tables to find the parent rev ID. If ar_page_id is NULL, I suppose it could guess passed on the archive table (title,time) alone.

@aaron , would like your help triaging this one.

In T425893#11955443, @HCoplin-WMF wrote:

Adding to @KineticPelagic 's comment -- could we get a #/% of files that have a number of edits that go beyond the threshold? I agree that we should not necessarily have an upper bound in place, but I would like to get a better sense of how commonly folks are running into this for prioritization purposes.

In T425893#12127513, @Urbanecm wrote:

However, do note that it's relatively straightforward to create such file (as an attacker), and this task removes the community's ability to handle such a file.

In T425893#12127484, @JTweed-WMF wrote:

How common is it for files to have a very large number of revisions?

This may come under Media Platform scope, but I have the same question as @HCoplin-WMF. How common is it for files to have a very large number of revisions?

Confirmed that this bug is fixed, so closing.

Based on the code flow here, this problem appears to have already been fixed:

		} else {
			if ( $onlyAuthor ) {
				$reason = wfMessage(
					'excontentauthor',
					'$1',
					$onlyAuthor
				)->inContentLanguage()->text();
			} else {
				$reason = wfMessage( 'excontent', '$1' )->inContentLanguage()->text();
			}
		}
		if ( $reason == '-' ) {
			// Allow these UI messages to be blanked out cleanly
			return '';
		}
		// Max content length = max comment length - length of the comment (excl. $1)
		$maxLength = CommentStore::COMMENT_CHARACTER_LIMIT - ( strlen( $reason ) - 2 );
		$text = $content ? $content->getTextForSummary( $maxLength ) : '';
		if ( $blank && !$text ) {
			// Don't display "content before blanking was ''" as misleading
			// This can happen if the content before blanking was two unclosed square brackets, for example
			// Do display `content was ""` if the page was always blank, though
			return false;
		}
		// Now replace the '$1' placeholder
		$reason = str_replace( '$1', $text, $reason );
		return $reason;
	}

I just tested this on testwiki, and it can be seen to work as expected.

https://en.wikipedia.org/wiki/Special:Undelete/Feted_Inner_Core shows the revision sizes instead of the byte changes:

Patches will wait for https://gerrit.wikimedia.org/r/c/mediawiki/core/+/1302993

Change #1303577 merged by jenkins-bot:

[mediawiki/core@master] lockmanager: avoid pointless retries in LockManager::lockByType

https://gerrit.wikimedia.org/r/1303577

As part of my Clinic Duty rotation, I keep this task with MW-Interfaces-Team because we are maintainers of Page deletion. I move the task to "Needs Further Discussion" because we need discussion on prioritization and on the unknown about whether the code in Translate can be migrated, too.