I have many disagreements with the dominant theory of raid design, one of which is gear-based alternative advancement that divides content into tiers based on arbitrary numbers rather that create fake difficulty. Whether or not you like synchronized dancing online, you should see something perverse in a system where everyone might perform all the steps perfectly but still fail the dance because they have not spent enough hours grinding gearscore.
The math on this gets interesting when the target gearscore is above what is possible as you start a new tier. This is one way to spread content over time. The first time you visit a new tier, let’s say that your gear plus tactics plus random rolls give you a 20% chance to beat the boss; the other 80% of the time, your tank is not geared enough to take back-to-back crits, your DPS does not down the boss before a lag spike leaves someone in a pit of fire, etc. An 80% chance of failure sounds pretty bad, but many players seem willing and able to accept it, especially given that you can try more than once per night, and you have a (1-0.8^5=) 67% chance to win before you wipe 5 times.
That sounds pretty standard for raiders. You wipe a few times, but you down the boss. As you practice and improve your gear, you wipe less, and then you get that boss on farm status. You might be performing exactly the same dance steps, but your gearscore is higher, so you no longer are facing a 4-to-1 odds for exactly the same sequence of button presses.
This strikes me as perverse and unsatisfying for the same reason that Desktop Dungeons did. You might have been a few percent more on-the-ball that time you won, but more likely the dice just fell slightly on your side of the margin of error. You get the illusion that you did it, yay team, but you are just grinding finite probabilities (or, when your gearscore is higher, grinding near certainties).
On the other hand, all my respect to those who overcome massive gaps in gearscore to succeed through perfect execution. The folks who are consistently getting server firsts a month before most people can beat phase 1 are not just hitting 5% chances consistently.