I'm very likely one of those players "on the cusp," so to speak. Few 3.5s trouble me, yet I trouble few 4.0s. My 4.0 matches have been competitive (certainly in actual "play" although perhaps not in score) and I don't feel any of the 4.0s felt "cheated" in having to play me when I was playing "up."
But going to a .25 rating system seems just a bit too much. We're blessed in our area to have lots of players and lots of leagues. A finer rating differentiation might work here, but in other places, where they sometimes have trouble fielding more than 2 or 3 teams per league/flight....eh? Besides, team/league play is where I've had the best chance to meet more people/players outside of my (playing) comfort zone....both higher and lower....in both age and skill level.
I wish the algorithm more appropriately rewarded an actual "win" and/or de-emphasized the so-called predicted result. I really don't think a top-end (say) 3.5 should have to destroy a low-end 3.5 -1 and -1. That process, in and of itself, sets up such vast differences within each NTRP.
A 3.5 masquerading around with a 4.0 mask on.