Preamble/tl;dr:
M+, specifically high-key m+, suffers from a lack of communication between the balance/design of affixes, classbalance and dungeon design/balance. This creates several issues, that I will talk about in detail below. Our fix for this imbalance was to delete keys and fish for new/better/easier ones. This solution to a problem was removed, but in a self-contained change that did not come with any major balance changes that addressed the underlying issues, which again shows a disconnect between different parts of a machine that need to work together for the machine to not break down.
This feedback is exclusively from a high-key pushing PoV, and some of these complaints do not apply to “lower” keys that are meant for casual play and/or PuGs. It is my personal opinion, albeit most people I’ve played with or used to play with have voiced similar concerns.
Imagine playing a game where you have to go through a gauntlet. People on the outside are supposed to make it harder for you to get there. Person A pushes a button and opens a trapdoor in the floor, so you would have to jump over the hole. Person B uses a lever that lowers the ceiling, so you would have to crawl for a bit. Individually, these slightly increase the difficulty. Combined, they create a scenario that is harder than the combination of the parts. That is what m+ feels like right now.
There has been many core-system changes since legion, some of which drove the “old” players away while bringing new ones into the fold who happily enjoy m+ now. I will try to look at most features without comparing it to the way it played in Legion, because there clearly was a drive to significantly alter the experience to make it more accessible for newer players. There will be the occasional reference to how things worked in legion though.
Reward structure/How to measure m+ success: While there was none in Legion, m+ still managed to attract a lot of people that immediately engaged with this new type of content, and it was hugely successful. An entire new sub-culture of players grew together into a large community, one that comfortably sits next to the raiding and PvP communities these days. All of this happened without any major rewards for doing so, which speaks of how enjoyable and fun m+ has been ever since its first iteration. This is something Blizzard can and should build on.
With the addition of a seasonal system similar to PvP seasons, as well as seasonal affixes, an official leaderboard on the blizzard page and even somewhat regular MDI events, everyone was hoping for actual in-game rewards, as well as better regulation/oversight of how success is measured and promoted for high-keys. Alleviating the need for third-party leaderboards like wowprogress or raider.io is something many people have been asking for, for a multitude of reasons.
Unfortunately, none of this has happened. Having this sort of functionality be official and accessible in-game, would be healthy for the community and the game. The current leaderboards have some major issues. Say if the same 5 people overtime a Tol Dagor 24, but then time a TD23. This new and potentially world-first run will not be shown on the leaderboards and the current understanding of almost the entire m+ community is that only in-time runs really matter. The current implementation of the leaderboards is simply not keeping up with the community perception of what constitutes the “best” run.
Now for the meat of this post: Balancing. It is obviously the main-issue and at the core of almost all issues that are being voiced, and it consists of 3 parts.
Affix balance: Affixes are the biggest variable when it comes to how difficult a specific key is on a week-to-week basis. Something people consider an easy key in this week, might be much harder the next, and even easier the week after. This has always been the case, even in Legion. Most of the affixes still feel like a 1.0 version that was introduced in legion, and did not get a tuning pass for the new expansion, creating multiple issues. Some of these have since been addressed, like bursting and grievous damage numbers, others have been changed to just work differently like quaking, but many affixes feel like they did not get looked at at all as to how they will work in tandem with class changes or the new design of BfA dungeons.
Class balance: M+ has always put a much bigger emphasis on utility than dps/hps, and while every classes potential in the four throughput-departments (single-target, cleave, burst aoe, sustained aoe) still play a big role, so does access to certain utility, both unique and non-unique one.
We currently see a much bigger emphasis on unique-utility than we did in legion - Shroud, Ring of Peace or Treants being some of the major ones.
There’s also non-unique utility like access to singletarget interrupts, which is why you see so many 2 melee lineups. Dispel-types for healers/hybrids or enrage-dispel would be another example of this.
Some of this class utility can now be accessed through professions, like group buffs for stamina, intellect and AP. Engineering now provides the ability to combat-rez, although the “backfired” version of this is very badly designed and needs to be looked at. Healthstones can be traded or accessed through the Proving Grounds.
Most utility however is still very much unique, both in design and who can access it, despite clear examples of potentially allowing for at least one more class or spec to have the same utility. For example it really wouldn’t hurt class fantasy or balance to allow mages to cast Mass Invisibility, same as rogues can use Shroud of Concealment. They already do in PvP.
Compared to Legion, most classes got their utility cut when it was used to aoe-interrupt mobs - one of the major changes to general gameplay was to cut down on just mass-interrupting large groups of monsters until they died.
Most AoE-CCs are now locked behind a talent, been put on a higher cooldown, or had their duration reduced. Some have been removed entirely. Not all classes have suffered in this way however, with certain specs/classes gaining more utility than they had before. This strikes everyone as odd, i.e. no one understands why monks now get access to both leg-sweep and ring of peace, while shadowpriests had their already weak cc-capabilities tuned down even harder. This might come on the back of PvP balancing, but obviously this feedback is solely concerned with how it impacts m+.
Dungeon balance, or rather design: The third part is the basic design of dungeons themselves, where we see the biggest difference when we compare BfA to Legion. Legion put a large emphasis on well-telegraphed and animated abilities that gave the player the time to react to whatever was happening, either though movement, or skill usage. Moving out of frontals/AoEs, chaining AoE interrupts, kiting, outranging, LoSing or simply using personal and external defensives at the right time were all critical to your groups success, and required a consistent level of communication inside the group. Certain dungeons required a specific amount of interrupts to lower the difficulty significantly or facilitate pulls that would ultimately save you time.
BfA has been a stark contrast in almost every single aspect of that. There are almost no one-shot mechanics anymore that require a players counterplay to survive. Most frontals, charges or AoEs are badly telegraphed, and mobs in general are smaller than they were in Legion, making them stand out less in all the AoE/animation clutter. Abilities that you used to outrange or LoS are now random and can either one-shot an unlucky player, or exist to disincentivize kiting. Most abilities just completely ignore LoS and go through walls.
There is very little consistency as to how these abilities behave. Two different mobs may use the same fixate ability, one deals physical damage, can be dispelled and stunned to remove it, while another cannot be removed in any way while also dealing magic damage. This is not intuitive at all, and explicitly counter to the point of making m+ more accessible.
Dungeons have more trash, and that trash has more abilities. Individually, none of the trash is as lethal as their legion counterparts, they do not one-shot and they deal a minor amount of damage. Stacked up, and scaled through the basic nature of m+, this starts to become a major issue on higher keys, as random, instant, melee-range only abilities start to melt tanks and melee alike.
Debuffs are incredibly inconsistent, and heavily favor poison-dispels over any other type. There is more lethal and dangerous poisons in dungeons than there is magic debuffs, the only type any healer-class can dispel.
This part of the “trifecta” also includes the zone-design. Dungeons are much more claustrophobic and limit the area you can fight in through space-restrictions and/or packing them with too much trash. Line of sight issues that lead mobs or bosses to teleporting the tank, and/or evading, and/or casting through walls. Objects like doors and gates that can be opened/closed are a part of this as well, vehicles that players can enter etc.
All of the major issues players have with specific trash mobs, affixes or dungeons seem to exist because it “feels” like these three core-systems have been redesigned and re-iterated on independently.
What this means is that certain affixes make dungeons much more punishing, both in general and specifically.
As an example: Bolstering/Grievous/Tyrannical in a dungeon like Waycrest Manor. There are many mobs with different levels of health. A single pull of 6 mobs can contain mobs with 4 different levels of health, which makes bolstering much harder to manage. This then ties into requiring much higher priority target DPS to even out those health-levels, which has been drastically reduced for most classes, while disincentivizing bigger pulls. This creates the need for certain classes.
Keeping mobs in place is much more important nowadays due to how most classes AoE has been limited in range/area or is based on a temporary ground-effect, and Stuns, Ring of Peace, Treants or strong slows all become much more potent but further limit the viability of classes at the high-end.
The most lethal/important mobs in these pulls are not the ones with the highest health or the biggest size, it’s the ones with the lowest health and smallest size. You’re basically playing whack-a-mole-the-nameplate when interrupting maggots, of which there are 2 types with the same name but different health.
Trying to gauge what direction Retch is being cast at is much harder when mobs are jumping to random players and are inside a dogpile of players, mobs and spell effects.
The amount of unavoidable group-damage severely strains the healer, who has less tools to deal with that damage. Healer-design changed compared to Legion, with all healers having less instant-/high healing abilities at their disposal, while grievous remained unchanged for months.
All of this is one pull, and all of these are individual mechanics that would be fine in a vacuum, but combined are much worse than each individual part. They’re not impossible to deal with, but they work together in ways that should be looked at more closely to prevent unfair situations from happening.
Lets look at it from another perspective by just thinking about one specific Dungeon: Tol Dagor. Designing this dungeon with m+ affixes in mind should immediately have led to questions and changes to the bosses, affixes, or the dungeon itself.
How to deal with necrotic on first boss? Well, it’s been fixed now, after ~6months of beta+live experience.
How will Spawns of Ghuun interact with doors/gates that can be opened? They just run through them, even though the player cannot. We just have to assume this is intended, because no change came, and Infested is gone in a few weeks. Will it be the same for the new seasonal affix? We don’t know, because we are not allowed to test it on the PTR.
Multi-z-level dungeons have always been finicky when it comes to knockback effects or AoEs, yet we have the same clipping issues with abilities pulling one floor below/above us, mobs clipping through walls and evading/resetting, pulling the entire dungeon, or bosses simply evading and resetting mid-fight because a single add spawned in a location where it could not hit the tank.
To come to a conclusion, these examples are meant to explain and illustrate how class, affix and dungeon balance and design work and interact with each other. In Legion, this was a very enjoyable experience and everything felt incredibly deliberate. Some of the BfA dungeons do Legion justice.
In most dungeons however, these individual parts work against each other to create an experience that can, at best, be described as annoying and at worst leads to people quitting over an incredibly aggravating, frustrating and unenjoyable experience.
These are not issues of the game being too hard, because dealing with these situations is not difficult from a mechanical point of view. They are annoyances that feel like they are meant to waste your time, on your goal towards timing the key. It is like the gauntlet I described earlier, that you try to get through, while the people on the sidelines try to make it as hard as possible for you by throwing all kinds of things at you.
But all of this changes the moment a new week starts. You get a new infest rotation, and you can suddenly do a bigger pull here and there, or kill this pack that you previously skipped. Different affixes that don’t interact horribly with the dungeon design, or scale beyond what the DPS, Tank or Healer are capable of, allow you to pull bigger and save time. What this inevitably leads to is being able to comfortably time a dungeon at +20 in one week, while barely timing it on 18 on another. And this is a huge problem.
Same as within a singular week you barely time Dungeon A on +18, while 2-chesting Dungeon B on a +20. Now you got that +22, but the new key is for Dungeon A so you have to deplete it to +18 before you have a realistic chance of timing it.
This is current high-end m+ in a nutshell, and the removal of being able to “fish” for new keys by deleting bad ones, while in itself a good change that is important for the health of the scene, ultimately made things even more frustrating while illustrating the same disconnect between seperate parts of the equation. It worked well in legion, because despite all complaints, there wasn’t a single dungeon you couldn’t at least time on +28, while really only one dungeon was timeable on +30, potentially a second one if we had more time with it. It forced people to engage with harder keys, learning new strategies, and overcoming the struggle that they previously thought was impossible. If you had 5 Seat keys, you were forced to do the dungeon, and you improved after slogging through it, whereas early BfA just led to people completely ignoring Tyrannical Temple or Underrot.
The change itself is also self-contained however, which is the core-principle that plagues most of m+ right now. It removes the players ability to game the system and forces players to engage with all dungeons, not just the easy ones. That is good.
But it does so without addressing the major balance concerns players have about dungeon difficulty, and more important the difference in difficulty, on a week-to-week basis. The “fix” for one of the problems Blizzard saw with the current way m+ is approached, suffers from the same issues as all the issues that lead to us utilizing this “fix” in the first place. This, at its core, is why the m+ system is seen as broken by a lot of people.
It’s a bit of a hit and miss if it works or doesn’t. Please let us make a macro to use that ability or something 
(maybe you can already but we haven’t worked it out yet)