In DCSS (and presumably other roguelikes), one of the most frequent complaints is on the difficulty of the early game. It is one I often complain about, though never in a public forum. I think it is too often mentioned but not backed up with any real substance. For example, someone might just say, "The early game is too hard!" I think it is easy to see why that isn't really constructive, but perhaps they go a step further and say, "The early game is too hard because of x,y,z - please remove x,y,z." But again, is that productive? I don't think so. I don't think Crawl would be a better game if the developers decided to remove Grinder, Sigmund, and Orc Priests altogether.
The mid game of Crawl (I would define "mid game" as immediately after Lair, though I'm tempted to include Lair as well) is one of the best, if not the absolute best roguelike when it comes to balance. Essentially every death can be traced back to a tactical or strategical mistake and the player has acquired enough tools to deal with almost every conceivable situation. A cautious and successful player (not usually me) will have enough raw power and escape mechanisms to either defeat or avoid whatever the game throws at him. I would guess the late game is similar, but I haven't made it there consistently enough to judge.
So, how does the balance of the mid game relate to the (perceived) imbalance of the early game? I think it is quite simple, really: when you die in either part of the game, it was almost certainly your fault. However, personally, when I die in the mid game, I can quickly assess what I did wrong and how I could have overcome that situation. When I die in the early game (especially on the first five floors or so), I find it very easy to blame the game, even though I'm subconsciously aware that it was probably my fault. The reason(s)? A lack of tools! In the mid game, it is not at all uncommon to have several scrolls of Blink and Teleport, potions of Heal Wounds, Speed, Brilliance, and others, and wands that inflict good damage or impair the target are very common now.
Major threats can appear as early as the first floor, though the second floor is much more common. While identifying all of your scrolls on the first floor is a common practice, many players will not have identified many potions. In addition, the practice of "read IDing" scrolls will often leave you on the second floor with no scrolls of Teleport or Blink. Running away is rarely as simple as it sounds, especially if you don't have access to those critical tools yet. And chugging potions randomly, hoping to get a good one to save you, is not an uncommon death sentence - slinging back a potion of Paralysis is a good way to doom a young character.
The early game does a lot of things right: it introduces curses, the food clock, many different types of items, and a nice variety of enemies to keep things interesting. One thing it also tries to teach is the practice of avoiding combat. I don't think this is really taught all that well - I've picked it up, slowly, but it isn't really as intuitive as the other "lessons of the early game".
So, how would I amend this issue? I'm not entirely sure. For example, removing branded weapons off enemies sounds like a possible solution, but that really just removes a potential reward for the player, which isn't a good idea at all. Ogres are interesting enemies because they're very dangerous in melee, but move slowly - but does that mean every dangerous enemy early on needs to move slowly? Probably not. Having different types of enemies is important, and that includes movement speed. Perhaps early uniques shouldn't be able to follow you up or down staircases? Again, a possible solution, but it would also teach something that wouldn't be true later on and also something that isn't always true - while "stairdancing" is a common practice, sometimes using the stairs is a good way to kill yourself in one action.
Perhaps the biggest indicator that the early game is not impossibly hard is that a few fantastic players have managed to obtain rather large win streaks. This wouldn't really be possible if the early game was entirely dominated by the random number generator, unless we want to dig into the argument that those players are just lucky (but that doesn't sound fair). Maybe part of it is investment - I'm a lot less interested in why I died on a character I've been playing for ten minutes, but I'd very much like to know what I could have done to save my five hour character. Ultimately, I think I can sympathize somewhat with the Crawl development core, as this issue, however frustrating it is for some players, is not one that is easy to solve, especially not if one wishes to maintain the "integrity" of the game, as it were.