1. Things setting up on choice-locked keldeo
My issue with your argument about keldeo bluffing is this: the ebelt set is great at bluffing as choiced, but the choiced sets don't bluff as ebelt. In 99% of games, if I send out keldeo against your +1 dragonite with multiscale broken, you're going to be perfectly right to conclude that it is scarfdeo. Furthermore, if that keldeo then proceeds to outspeed my dragonite and KO it with hp ice, i'm going to go straight to my SDzor and set up. Bluffing works great in the other direction: for example, if I hydro pump the blissey switch-in and it does 30%, and it has 29% health left now, and i hydro pump again instead of secret swording, then maybe the next time celebi comes in on my hydro i can hp bug it. Your argument would only really make sense if choice sets were rarely used, when in fact 54% of keldeo are choiced (according to most recent stats) and only 14% are ebelt.
My point was basically this: given that keldeo benefits so much from being choiced, and that so many of the pro-ban arguments depend upon the boosts it receives from scarf or specs (it's nowhere near as difficult to counter if we ignore specs, and doesn't function as a revenge-killer / late-game cleaner at all without a scarf), we have to look at the fact that it's one of the most set-up-on conducive choice-users out there--secret sword has a type immune to it, as does hpump via water absorb, dry skin, etc. It doesn't get decently powered coverage moves (it's the only OU mon I can think of that regularly uses a 50 BP move--horrible movepool in terms of coverage). Furthermore, its STAB's, while powerful, have meh coverage--and what's worse, one of the most major types that often resists both are dragons, which happen to be some of the best set-up sweepers in OU (dragonite, mence, lati@s, etc.). And sure, it can revenge them with hp ice on the scarf set, but hey, steels, another hugely threatening type, can set up on that! (Non to mention that half the set-up sweepers in OU can take many scarf hp ices from keldeo because it's so weak). Compared to something like terrakion, whose SE has literally no immunes and few resists, and who doesnt need coverage moves--basically literally, ever--keldeo is so much more prone to letting sweepers set-up on choice-locked variants of it.
2. Pursuit-trapping
The standard jellicent listed first in the smogon analysis has 44 Spe EVs, specifically to outspeed 8 speed scizor. Not only is this totally logical (given that checking scizor is a big perk of jellicent's), but jellicent's taunt is one of its greatest assets, and it needs speed to make use of it (taunting politoed before it toxic, ferro before it leeches, etc.). Outspeeding standard bandtar requires 88 spe EVs, which comes out to 11 points less in a defense--generally not something that will often cost you a game. Again, because jelli can make good use of the speed, this is a totally viable option. Just because these are not super common (if indeed they are not--I don't think I know a single skilled played who has run anything below 44 speed on jellicent) does not decrease their effectiveness. Meanwhile, I don't really see BP being such a big hindrance to celebi--you can absolutely still run psong, as very often you'll be needing them independently of each other, so they wont conflict. I dont really see why u-turn is a big loss--celebi never used it for anything but the free switch, and it's actually nice quite often not to take the iron barbs / rough skin damage.
And so what if these changes are being made to escape from pursuit (although I'm not even saying that they are 100% at all--as I said, jellicent makes great use of the extra speed for the faster taunt, and celebi runs u-turn 33% of the time anyway--and interestingly, psong only 8%--and basically doesn't lose anything by using BP over it)? I could make the argument that ebelt keldeo's HP bug is almost exclusively for celebi. So? Mons run unusual moves to get by their counters all the time, and there's nothing wrong with it. Celebi is one of keldeo's best and most common checks, so keldeo runs a move to beat it. Similarly, ttar and scizor are celebi's greatest enemies, so it runs a move to escape from them. (Not to mention that celebi gets far greater use out of baton pass outside of escaping pursuiters than keldeo gets from hp bug outside of beating celebi.) In fact, being able to use a wide variety of moves viably to get past counters is a huge bonus to a pokemon, and is one of the things that people are claiming makes keldeo broken!
3. Weakness to passive damage
You're right, you could say this about most any pokemon. I was more expressing that keldeo is not like tornadus-t in its ability to beat its counters by outliving them via regenerator. I agree with you, keldeo is probably less prone to being worn down than most pokemon due to the SR resist (equal to garchomp, yep), but it's can't play the "spam attacks and then switch out to wear down counters while staying in perfect shape regardless of residual damage" game like torny-t did. Not an important point, was just noting a significant difference between the two mons (particularly important since I'm pretty sure that was the deciding factor for a huge portion of the voters).
4. Checks have reliable recovery
I'm not 100% sure what you're arguing here. I agree that something being able to recover does not make it a foolproof counter, and that yes, recovering does give the opp a free turn. But just because something has a recovery move doesn't mean that it has to use it every time it takes damage. Your assertion about the momentum gained is basically just saying "if I can 3HKO your check if I start with an obvious move, i don't lose any momentum by using it, because they are forced to recover or risk the 2HKO the next time I'm in". Absolutely true. But this has nothing to do with them having recovery--it has to do with the fact that keldeo is a potent attacker who can 3HKO even its checks with proper prediction. I was making a totally separate point.
Which is that, as the converse to the point above, keldeo can't just wear its counter down over the course of the game. Landorus-t may be a more solid check to terrakion than celebi is to keldeo (no terrakion--bar hp ice lol--can come close to 2HKOing landy-t, whereas keldeo easily 2HKOs with hp bug), but if your keldeo does not have hp bug, you will NEVER get past that celebi without help from other mons or hax. It can take ANY 2 attacks from full health, and heal off more damage than you deal. On the other hand, the terrakion user can and will often get by that landorus-t through smart play--every time I stone edge as you switch in, I'm 30% closer to breaking though. If I switch out each time to something that forces landy-t out, it only takes 2 hits before terrakion can break through landy the next time I send it in. Meanwhile, this will literally NEVER happen with keldeo, because each and every one of its checks has the ability to recover off damage as fast as keldeo can deal it (basically).
5. Not over-centralizing
My point was mostly about the fact that obscure keldeo checks have not become common in OU (slowking is #132, for example). Not a big point, but the fact that rotom-h rose a ton in usage in genesect's prime but that slowking, which is a much more viable mon in terms of stats and movepool (and which shares typing with bro, which is relatively commonly used), has not, says a fair amount amout how much keldeo really has shaped the meta.
Don't have time to reply to the last one, but I will in a bit. Interested to hear your thoughts, myzo!