The only problem with occasional forced deaths from e.g. traps is that the community mismeasures skill by streak length. If there were a way to declare the scoring objective of a game in advance, streaks wouldn't matter, winrate would. Declare at the start whether it's turncount, realtime, win-rate, 3-rune, 15-rune, etc. This would allow elite players to differentiate themselves by win-rate on difficult builds or other types of challenges, where the win-rate isn't absurdly high. Then it would not be time-prohibitive to compete in Crawl, and the meta would evolve as a result of accurate feedback.

This feature should be programmed into online play, but a hacky individual implementation would be to record the game scoring objective in the in-game notes before playing the first turn. Which is still not ideal, since the player has seen his starting vault, and it's difficult to datamine and prevent cheating. But it would be something.

I suspect Crawl "wants" to become more difficult, but is held back because elite players insist on protecting their ability to streak, which is the best way they have to compete currently. They rightly hate losing a streak to BS RNG, because it means more pointless grind. The situation generates needless frustration for everyone.

How fun can it be for elite players to be able to endlessly streak the optimal build for each species? Most alternate builds just want to converge on the optimal after a rought start anyway. It encourages skilled players to drop out after achieving enough skill to reach the competitive grind stage.

It would unlock a lot of strategic depth to program the online starting interface so that for example this were a competitive genre: narrowed and/or randomized choice of god, background and species, with god conduct enforced in-game. I don't know exactly the best way to do it.

If elite players aren't forced to play quickly to mine ridiculously long streaks or compete in grindey tournaments, then the game can evolve in a more chess-like direction, with less tab-spam and more engagements that matter. Even the realtime speedrunners will benefit from the improved meta evolution of the slow-play sector.

Streak-mining grinding should not be a competitive category. It is like awarding Olympic medals to the guy who has run the most consecutive career 4-minute miles. You tripped, start over! That's not a thing.

Lastly, to annoy everyone, skilling without reference to marginal cost is like shopping without reference to price.