Yesterday I advocated returning to something closer to linear progression for raiding and heroics: raid X, then raid Y, then raid Z. The goal, and hopefully the effect of it, is to separate players out by skill and experience, so that if we are in a certain tier of content, we are with similarly skilled and geared players. Contrast this with heroics which throw everyone into the same group and raids which divide only into normal and heroic.
There are some major problems with a linear progression design.
One problem is the bottleneck guild-killer boss. Blackwing Lair was notorious for these. One boss that players just could not get down. After months of no progress players would quit, sometimes the guild, sometimes WoW altogether. Beside this, getting stuck with one boss as the only new one to try isn't fun at all. Variety is important and you can only wipe on Razorgore for so long before you get sick of the sight of eggs.
Don't have a single raid at each tier. Have two at least. Early Lich King had three different raids at the first tier, which while two were single-boss raids, they still served to give someplace different to wipe in, a needed change of scenery. These may also allow players to gear up a little more, reducing the relative difficulty of the boss they're stuck on and allowing for more progress. The extend to which one raid can make the other easier is important to consider, since the purpose of the other raids are not to let players skip difficulty content, but to give variety and some ability to self-determine the difficulty, without trivializing encounters.
Another way to bypass the bottleneck boss is to have a "one free kill" mechanic. Each week you get one free boss dead. It gives no loot, no achievement or meta-achievement progress, no quest progress, cannot be skinned, mined, or engineered. It's just not in your way anymore. Go on and die to something else. I'm not as enamored this fix, but I think it's worth thinking about, if only for the value in finding why it's bad.
Ideally every boss would be harder than the one before it, so that bottleneck bosses would be a measure of the skill of the guild/raid, with progress and skill being exactly linked. But skill isn't a number to be easily expressed as greater or less than a threshold. This means that there will always be earlier bosses which for particular raids are harder than later bosses, meaning that there will always be bottlenecks.
New players may find that the majority of the players have moved on past their tier. If they cannot find anyone to play with, they cannot advance, and so are stuck in limbo: lacking the gear and experience to progress, but unable to get the gear and experience if they're alone. We saw with in Lich King heroics when players had briefly gotten plenty of badge gear and had long outgrown heroic level gear, leaving alts and new players with very few people to play with. Blizzard's solution was badge loot inflation, driving raiders back into heroics, with what I believe were negative social consequences, mixing players with radically different expectations of performance and behavior. Now WoW has cross-server group formation, which I believe may increase the pool enough to address this concern.
There remains the rerolling problem. Switching classes can put an otherwise experienced player right back at the bottom for gear. While cross-server can help with the population problem, there are other problems. First off, being left behind means repeating all their previous gearing up, repeating content which may already be old. This process can be very slow, particularly if the gearing process means finding raids, meaning guilds, which the rerolling player intends to leave the moment they can. Possibly worse is that thus process means being unable to play with friends and guild members, which presumably was the reason for the reroll in the first place. For this problem I suggest a retirement system which would allow a character to be converted to another class, with some form of hastened leveling and gear conversion. Retirement would be a serious matter due to the cost of a character, but would allow much faster catch-up without making class something to change on a whim.
There is a third group: the players whose guilds exploded. They're not new players, so slogging through again may not be fun. They're not alts with a guild to catch up to, so the retirement system doesn't make sense, beside that they're not necessarily undergeared, just out of the system. These should be highly-desirable players: experienced, geared, and the last to give up (as opposed to those who quit after the first week of wiping), but they're not easy to find. A tool to help these players better find guilds, and guilds to find them, could help. Maybe it even already exists and is underutilized, but an armory tie-in would obviously be part of it.
Players Missing Content
There's an ongoing debate between "I pay $15 a month I should get to see content" vs. "why should noobs get everything?"
Vanilla used the solution which I will refer to as "lolwut?" Essentially no one did anything to fix it and maybe didn't see any problem. Burning Crusade used PvP and badge gear to help boost players up, which might have worked if not for the guild-wrecking 10 to 25 player raid transition. On the other hand, we ended up with "welfare epics". Despite being a retarded name, it somehow stuck. Lich King expanded the badge system further to give pieces of tier sets as well as badge rewards being upgraded with each new raid. This resulted in a whole lot of players in strangely high gear relative to their experience and skill, a recipe for disaster. This model also resulted in the opposite problem of new players missing the old raids as they became obsolete.
For this I have no good solution. Making the content easier ruins the fun of those already in it. Adding gear can ruin the focus of the game, turning it from experience-driven (meaning player experience, not experience points) to loot-driven, which can ruin it as badly as not seeing content. Flattening all raids to the same gear level would cause rapid burning of content, followed by a whole lot of bored raiders, and then the next raids may be too easy. Exclusive content sounds great, but often the excluded don't like it much.
I probably missed some problems, but those are the biggest three.