"Don't say something mean" is a lower bar than "say something substantial", which is what your post said the standard of endorsement should be. There's going to be a mental tax involved, asking "would PG be OK with me endorsing this? is my threshold for 'substantial' higher or lower than his?" to each comment, weighing the risk of losing the privilege for guessing wrong. I've apparently already mis-judged what you think warrants endorsement. The meta-moderation aspect of someone subjectively reviewing my endorsements is a disincentive to participating in that moderation.
The number of people that can upvote is very high, yet most comments probably sit at 1 point despite not being mean or stupid. Now restrict that pool to 1000+ karma users, add a mental tax to the task, how many of those same comments are going to get two people to click "endorse" if they couldn't get one person to click "up"? I wouldn't put money on the rapid-fire sub-second moderation materializing, especially outside the top 10 stories or at odd hours.
Wait, does this mean that the motivation for filtering all comments is to combat comments that should already be flagged?
I'm getting a stronger and stronger feeling that there is plenty of room for focusing the stated goal of this change, and then perhaps redesign the solution.
Assuming this is how the endorsers actually behave. There's a good chance that they will end up acting more as 'curators' than endorsers, and only treat the endorse button as a 'like' button.
Maybe the verb 'to endorse' is a little bit too strong then. How about a button that literally says 'not stupid or mean' or 'appropriate', so that one doesn't have to indentify with that action as much.