The public git repository was created this year I think and it is announced in red on http://www.nethack.org/ with two links to http://www.nethack.org/common/git.html
I suspect Nethack transitionied to git a year ago (from svn or possibly directly from CSV).
"The license was written in 1989 by Mike Stephenson, who patterned it after the GNU bison license."
https://en.wikipedia.org/wiki/NetHack#NetHack_General_Public...
Another project with similar policy is OMNeT++ [0]. Simulation models are developed openly in the public [1], but the development of the kernel is done by closed group. Well, they also have restrictive license, forbidding commercial use, but what I am talking about is development model.
If you look at UnNetHack repository [2], it is forked from unofficial "Vanilla NetHack" repository [3] that contains only 53 commits, one for each released tarball. Now, with access to official git repository, it will be possible to track the changes between releases.
[1] https://inet.omnetpp.org/Repository.html
Having played a lot of both NetHack and Crawl, I think it's pretty bold to claim either game as the "best game ever". Both suffer from the same problem: past a certain point, your character has too many options for every situation and choices become boring and trite.
Crawl tries to "fix" the problem by making a lot of your powerful options unreliable. I believe this is the wrong way to do it. It messes with the player's natural feedback loops by inviting them past a point of no return demarcated by a hidden variable (the random seed governing the unreliable option).
Both games also suffer for the fact that they misuse a food clock. A food clock is a crude mechanism to stop the player from abusing regeneration of health and respawning of monsters. The problem with it is that it's a linear mechanism which simply does not function properly with a nonlinear game. Again, it invites players past a point of no return in cases such as foodless dungeon branches. Crawl tries to "fix" this by giving the player way too much food, but that just lengthens the feedback loop without actually solving the problem. The original Rogue (and its spiritual successor, Brogue) has a working food clock because it's a strictly linear game with a finite amount of food balanced against a finite number of areas to explore.
Both Crawl and NetHack also suffer heavily from the problem of "windshield monsters". Like so many bugs splatting on the windshield, the games lull you into a sense of complacency until inevitably you walk around a corner and something really nasty notices you and you die. Yet another case of a long feedback loop that messes with the player's expectations. This is the game play analogue of jumpscares from the horror genre; a cheap trick that ends up pissing off the player most times.
That just means you have to be careful so that can never happen. For some that makes the game boring, for others it's part of the fun. I like being prepared, so it's fun for me.
The boring part of the game is not the nasty thing that killed you, it's the thousand useless orcs you had to kill before it happened.
DCSS does suffer from the optimum local minima thing as well, though. There are a lot of big design choices that stick around because they've always been there, rather than being good choices. For instance: the difficulty curve, melee being superior to spell casting, big random damage spikes, length of the game (too long), tediousness of summons / spell casting, kiting as the optimal strategy 99% of the time, and the other things that make dcss what it is.
I'll say this: I've played both NetHack and Crawl (and many other roguelikes) and the main difference between the two of them, for me, is that I have fun playing Crawl and I have never enjoyed myself playing NetHack.
For example, I like the fact that almost every editor on my mac obeys the standard shortcut keys (at the minimum, apple+x,c,v for cut/copy/paste, apple+z undo, apple+s save, apple+q quit, apple+f find, apple+g find next, apple+a select all).
If you started vim or emacs today, would you choose on a mac to not follow the standard interface rules, when the editor was first started? I would (personally) hope not.
* Well, not all of them, save doesn't really make sense for a terminal, and 'cut' can't really be implemented for a general terminal app. but copy/paste/find/find next/close window/select all and others are all fine).
With commercial products it is more common for products to die, not be upgraded, or have significant major upgrades. With open source instead the same products (nethack, vim, emacs and bash come to mind) can live in forever, gaining small improvements over time.
On one hand, this is great! You have a long-lasting product you can learn and use for years and years. On the other hand, I don't think nethack is the perfect game, vim the perfect editor or bash the perfect shell, they are very hard to displace.