Chiming in with some quick thoughts here as a council member who's been in favour of Sleep Moves Clause > Sleep Clause for years now, while also thinking it's a much better change for this gen than previous generations.
1) For me there's been different standout offenders with sleep over the course of the gen, especially Red Card Amoon early on as well as cheesy stuff like SD Hypno Val, but for the current meta the big 3 for broken sleep are Hypno Rai, CM Hypno Val (with Hex Tera Ghost), and Lilligant-H on sun teams. All three of these mons can run away with games with a little bit of luck (40 percent to get at least one free turn with Hypnosis, and 50 percent for Sleep Powder). These odds are too low to claim consistency from the users' perspective, but still higher than everything else that would be considered "hax" from an opponents' perspective.
2) The counterplay spread for sleep users is pretty poor. The 3 sleep users I mentioned are very fast, especially the latter two which are often faster than entire teams. Direct counterplay to the sleep move is pretty rare outside of Ghold/Glisc, and both are pretty bad into Rai, + increasingly poor into the ever-widening range of offensive threats that we've been getting (including Wake/Gouging Fire that are found on Sun with Lilligant). Everything else is pretty limited in terms of proper use cases with teams that dont make the tech into a liability.
3) Sleep sacking is usually a bad response to these threats. In context, you have to remember that these mons are being paired with some of the most powerful offensive threats ever made, and the expected response with sleep sacking is to go into one of your mons (often u dont really get to choose much because this has to be at least a passable switchin), sack it to sleep, switch out IMMEDIATELY, and then hope you brought a second switchin, otherwise you're gonna be going to a third mon and probably taking damage to revenge kill. Even in the better version of this scenario, you've left a mon on 0 sleep turns and the sleep mon can still double or just switch out. This is especially true of Lilli-H, who is very useful as speed control for sun builds (actually a top tier style!) and has fantastic natural coverage thanks to the recent DLC where it gained Triple Axel. The issues with sleep sacking also come into play when combined with the odds calculations mentioned in the first bullet point - should I be going to Clefable to revenge Lilligant-Hisui if there's a 50 percent chance that this play will sack my Clefable for no reason? Should I have cleared the high building bar of always double-layering Lilligant checks/always carrying a priority that can kill Lilligant (it could still tera)? Is it even an affordable risk to be making decisions like this when Walking Wake and Gouging Fire are warming up to knock your socks off?
4) Following on from the last point, the ineffectiveness of sleep clause has been largely covered up by the fact that the only sleep users in past gens were either mons that were awful outside of sleep (Breloom), mons that do their job perfectly fine without having to risk putting anyone to sleep (Gen 8 Blacephalon), or defensive Grass types that generally had a wider range of mid-ground switchins (and therefore a wider choice of sleep sacks for the opponent + an immediate passivity issue that occurs once they've blown their sleep). Now that we have mid-level fast offensive mons that can run sleep as part of a decent set, it becomes more clear that the nerf we implemented on sleep all those years ago may not have been as universally effective as we thought.
And before anyone says "shouldnt have freed Darkrai", we had a solid two weeks where no one on high ladder was using HypnoRai and what we learned was that Darkrai is roughly the 7th best Dark type in OU, somewhere around B rank viability, and has some actual utility to offer to the tier. The fact that there's a jump from "balanced and B rank" to "wait should we ban this" off of one move is perhaps a sign that we underestimated how good sleep was even in a Sleep Clause metagame.
People can say "it's not sleep it's Rai" all they want, but I can say "it's not Rai it's sleep" with just as much validity. Unlike most debates between banning a pokemon and banning a broken element of said pokemon, the non-pokemon ban actually has a good tiering argument behind it. The only reason we're in this position in the first place is because we decided to nerf a broken mechanic instead of ban it, something we'd never do in this decade. All of this is before we even consider the way sleep is used on multiple pokemon right now.
5) Sleep fucks with the balance of Gen 9 specifically. Despite the insanely power-crept mons that we have been receiving all generation, and the crazy concept of Tera, this meta has (mostly) functioned well on a competitive level. This is largely down to a few interactions that do just about enough to keep the meta honest, instead of letting people succeed off of randomness. Two of the main ones that we see all the time are Tera4Tera and Kingambit revenge killing. Essentially the idea here is that if you load one of the many super fast and super powerful mons that we have access to, and then use the Tera button to smash through something that normally checks it, you can then mark down that your opponent can't Tera anymore, and cleanly respond to that Tera threat with one of your own Tera types, or a boosted Kingambit Sucker Punch. Most of the same concepts apply when you run into a threat that is fast and breaks through everything without Teraing - you use defensive Tera or Kingambit to stop the threat, live to fight another day, and then hope to win now that you're playing against things that dont beat ur whole team. Unfortunately, fast strong sleep users can kinda bypass these aspects of the metagame - defensive Teraing or Sucker Punching into a sleep move is deeply unappealing, and in this case you're often banking on them straight up missing the sleep move.
For an example of this, imagine a Kingambit trying to revenge kill a +1 Ghost Tera'd Valiant (pretty reasonable situation). The Kingambit player has an option to either Sucker Punch raw (and risk Vacuum Wave or Encore), or make the safer play and Tera to live all moves and KO with Kowtow Cleave. If the Valiant player is using Vacuum Wave, they will probably click that and do a small amount of damage if the Kingambit Teras. If the Valiant player is using Encore, they have a 50/50 choice to make between attacking for chip as Gambit Kowtows or predicting the Gambit Sucker with CM -> Encore. If the Valiant player is HypnoHex, most of the playing aspects on both sides get replaced with an RNG check - Hypnosis covers Gambit Sucker Punching, and 60 percent of the time it will cover Gambit Teraing. If the Gambit makes the safe play that covers the most options (Tera Kowtow), they still need to wake up instantly to not die to +1 Ghost Tera Hex. Note that in this situation, while the HypnoVal user does have a higher chance of coming away with absolutely nothing (miss Hypno and die), they also have a higher chance of coming away with everything (dead Tera Gambit). On top of that, they've managed to achieve all of this without making a decision on their end beyond the preset "I Ghost Tera and then click Hypnosis when Gambit tries to revenge me". This type of situation also applies with Gambit vs Lilligant, and it also applies with other defensive Tera mons vs Darkrai. If you defensive Tera your Gking vs Fairy Rai, then you have 80 percent odds to either dodge the Hypnosis or wake up in time to get a big hit off. This sounds great! until you realise the other 20 percent of outcomes is u losing ur Tera + Darkrai check and then probably losing the game afterwards. In these cases the sleep user is effectively saying they're not gonna try and break their opponent down through any normal, non-RNG reliant means, and instead opt for a choice that will take the outcomes out of both players' hands. This was always a pretty risky thing to allow.
6) Sleep fucks with the game in general. Something that is pretty much universal across almost all metagames is the idea that offensive teams need to be able to pile pressure onto specific defensive staples as a way to overwhelm them and get a winning position, while defensive teams want to cover threats and have a backup plan where possible against top tier threats. Mons that use setup + sleep can bypass this to an extent, since you can play a sequence of clicking setup as they go to their first answer -> clicking sleep to decommission that answer -> hitting the second switchin with a boosted move. The result of this is that instead of working off the normal mons fundamental where a mon just has to avoid a 2HKO and retaliate well in order to be considered a defensive answer, you're now left with a much more difficult standard of having to switch into a pokemon that already got a free turn to boost.
This becomes relevant when preparing for Darkrai with a balance or BO team, because a large portion of otherwise fine Darkrai counterplay is left to either risk the sleep rolls or try to sleep sack -> actual sack to get the second answer in without dying. After turd2's Darkrai rampage about a week ago, I decided to look into what defensive cores could work against HypnoRai reliably, and there are some decent balance cores that can answer it (normally involving some sort of Spdef water + Unaware + Zamazenta complex), but it is very resource-intensive and not particularly friendly to finding new structures. Outside of these structures, you can still do some things in-game to force Darkrai to hit its moves as much as possible, but this is only for creating odds, not for creating clear paths - without pretty specific counterplay, you can still get posterized by turd2/Ojama/Vert with some slightly bad luck. As I mentioned in previous posts on the metagame thread, I do think it's good to have some more healthy mons that can make balance teams sweat a bit, but in order for this to actually help balance the metagame, you need to be able to link good performances from those mons back to good decisions in the builder or in-game. Adding an RNG check to using and playing against balance helps nothing - it only hurts good players no matter what side of the game they're on.
While some aspects of this dynamic can be cool - when Vert uses sleep mons he'll always try to consider whether it's actually worth it to click the sleep move overall - there will always be players that mostly aim to click the funny button and see how far their Darkrai can go. This approach is brainless enough that even the weak (less than 50%) odds of getting at least one turn of sleep start to become unfair for other players to deal with.
7) I'm suggesting a sleep ban because it's an easy, tiering-friendly way to address at least some of the meta-related BS that's going on right now and the arguments against it are less convincing to me than the arguments against the mons we could target right now. This does not mean I'm opposed to looking into other aspects of the meta - I think this meta is disgusting as hell right now! I don't think banning sleep will magically fix all of that! However, parsing through which mons are broken and which mons are just really strong in such a power crept meta with confounding factors like Tera is an endeavour that will take some time, and unlike previous DLCs we do not get an easy way to reset these decisions if it turns out we made a mistake. So yes, in a sense it isnt coincidental that I'm bringing up sleep a couple weeks after DLC 2's release - I think that when there's a good decision available we should take the opportunity to make it, cause lord knows that tiering this meta isn't gonna get any simpler in the near future.