he still remains a god whose only drawback is the huge amount of time you spend juggling your inventory while rewarding you with a lot of duplicated powers that only accomplish letting you train one skill instead of numerous skills
no, the main drawback of Nemelex, like
every other god except chei and xom (and maybe Trog, I guess) is that you cannot worship a different god.
In addition, Nemelex does absolutely nothing for quite some time upon joining.
Nemelex has problems, but his power level is not anywhere near the most pressing.
---
I'm going to ignore deck of wonders in this next part, because I think giving Nemelex a strategic impact on the game is still wrong and bad and should be removed. Let's assume he gifts destruction, summoning, and escape only.
This post will attempt to explain from my personal perspective the good and bad parts of Nemelex and argue for at least keeping a god with the same general design in crawl.
---
Here is the design of Nemelex: he is one of few gods centered around active abilities. In fact, I suggest that Nemelex is the most active god; among other gods who have only active effects, we have Elyvilon, and I suppose also Zin and maybe you want to count Fedhas. I do not think it is a stretch to say that Nemelex feels more of an active god than Zin (none of Zin's abilities give you an effective way to kill enemies on their own) or Fedhas (you still end up fighting alongside your plants and doing much of the work yourself). Elyvilon is different in that Ely has a very narrow impact on the game, though that narrow impact is even more powerful in its own way than Nemelex (instakill offense beats even legendary hammer card, and guaranteed life saving and tons of self-healing beats out escape decks).
If you think that Elyvilon alone fills the niche of "active ability god" sufficiently, then removing Nemelex makes sense. I do not think this is the case (and Ely, of course, has her own list of problems, and personally I think they are even worse than Nemelex's since they are harder to solve, though that is of course a different topic so I will say no more here).
So what Nemlex does, as a god, is he gives you active abilities that dramatically change the way your character approaches dealing with enemies. Unless you choose to make pacification your main offense with Ely, no other god does anything like this. In addition, and importantly, Nemelex gives a tremendous breadth of different effects to the player--and this is something that definitely no other god can match.
The problems with Nemelex are, I think, almost entirely in implementation instead of design.
I will begin by discussing design.
Nemelex is a god based on giving the player a large selection of possible "effects". In the
particular implementation that is in crawl, Nemelex does this via gifting decks. In
general there is no reason a Nemelex-like god who provides a large selection of effects to the player must use decks, but I do not see a problem with implementing a Nemelex-like god using a deck-like separation of effects.
As I mentioned above, I definitely do not think that the breadth of effects that Nemelex offers (i.e. the number of distinct cards he gifts) is a problem. It certainly makes Nemelex more complicated and harder to learn, but that's ok, because that's the entire point of the god. In creating a god who provides a large number of effects, there must be some form in which these effects are grouped: if you do no grouping at all (or put every card into a single deck), you have complete chaos (like Xom) since the player has no real control over what type of effect he is attempting to use. If you go to the other extreme and allow the player to choose from every effect at any time (like with most Crawl gods) then you either must limit the number of effects pretty seriously or bloat the "a"bilities screen to unusability, as well as giving the player too many options at once.
So grouping the different "effects"--or cards, in Nemelex's case--into a few different distinct groups--decks--makes sense. Providing one single deck does not make much sense as I said above, so the minimum number that makes sense is two. Nemelex provides three, currently (plus the one (wonders) I'm purposely ignoring). The groupings should of course be cards that loosely fall into a single category; at the simplest, these categories would be "offense" or "defense". With Nemelex, of course, the categories are "destruction", "summoning", and "escape".
From here I will be using "effect" and "card" interchangeably; and "group" and "deck" interchangeably.
Now, how do you let the player actually access these effects? The simplest possible implementation is to just give the player a random effect from the group he chooses--or, exactly how decks work in crawl when you wield them and e"v"oke. But allowing only that is not terribly much fun: see Makhleb's major destruction, for instance, already does this. So now let's provide the player with some other ways of drawing cards.
One natural idea for an ability is to allow the player a bit of choice over which particular effect to use. Nemelex does this with Triple Draw. Obviously this should have a cost, and indeed it does: it both costs piety and cards. Stack Five is the same general type of ability.
Deal Four is a different idea--instead of giving the player more choice, you make the effect more powerful (but keep it random--or, in this case, make it more random!). This is a neat idea. It is not one that is necessary for Nemelex to exist, like I would suggest Draw One and some ability in the Triple Draw/Stack Five vein is, but it's certainly fun. This brings us up to, broadly, three different ways to access the effects that Nemelex provides: the simple, random draw; the choice-at-some-cost option; and the power-at-some-cost option. That seems pretty reasonable and not too complex.
So that's my view of Nemelex's design. I don't see large problems there, personally. Now some discussion of how Nemelex is
actually implemented in crawl, which is where the problems lie:
Nemelex is unusual in that he has two different sorts of "piety". There is the piety that the game actually calls "piety" and that you spend on his abilities, and then there is also "piety" in that what you actually spend to access effects are
cards. This is needlessly complicated, and I'm sure it stems from implementing Nemelex specifically as a card-god instead of as an effects-god (and it is Nemelex the effects-god that I am suggesting is worth keeping, remember). Since you get both "piety" and "cards" by sacrificing items to Nemelex I hope it's clear why I would say they are really just two different forms of piety. I'm not sure that this is a bad thing (and if Nemelex is to remain the card-god then keeping the two separate types of piety is pretty much necessary).
I will now use the term "piety" only for the thing that crawl calls "piety" from here on.
Piety gain from sacrificing items is tedious, of course, but there is, thankfully, an easy fix: now that the player no longer has any control over which decks Nemelex gifts, switching to piety-on-kills is, from a gameplay standpoint, largely the same as piety-from-sacrifices.
The implementation of decks is also, obviously, problematic. It is probably Nemelex's largest problem, since it turns out that crawl decks not only have different cards but also different power levels: plain, ornate, and legendary. So actually Nemelex gifts not three but
nine decks, and then additionally you get multiple copies of each deck (plus three more for wonders but again I am purposely ignoring that it exists).
One possible solution is to remove the different deck power levels entirely (so there are no longer "plain", "ornate", and "legendary" decks but just "decks"), and then to automatically merge decks of the same type together when the player acquires multiples. Think of it like ammo: you have 10 arrows, and you pick up 5 more arrows, so now you have 15 arrows. Similarly you could have 10 cards of destruction, and pick up 5 more to have 15 cards of destruction (the actual number of cards could be hidden from the player or not). This, of course, would require changing Stack Five to some other ability--or removing it entirely (since it is the same general ability as Triple Drawk the latter is not too problematic).
Those are the two interface problems that I see mentioned in regard to Nemelex. Condensing decks does not entirely eliminate Nemelex making inventory management more ... time-consuming (? this is not the word I want but I think my meaning here is clear) since three decks still occupy three inventory slots, but three is much less than ten or fifteen, and my other ideas (such as making decks into virtual items that you would view via the "a" or "^" screens or something) have other problems.
Nemelex specifically using evocations instead of invocations is again an artefact of implementing Nemelex as a "card-god"; if he stays the card god then this should stay but there's no need for a Nemelex-like god to use evocations instead of invocations.
Finally, there is, of course, how powerful Nemelex is as a god. There are many different ways to tweak this if it's necessary (I will make no arguments either way here, since I do not think I can do so adequately). You can tweak: how many cards he gifts, how powerful cards are, how cards scale with evocations, how much Nemelex makes cards better than not-Nemelex (including not making them better at all!), etc.
Since Nemelex is already implemented in crawl in the way he exists, I think that instead of scrapping Nemelex entirely and starting over, it makes more sense to take Nemelex and improve what is already there. But I have not seen the code; perhaps this is not, in fact, the way to go.