If hell were to be upgraded at all, I would suggest nothing about the aesthetics, but revamping the AI to make it a significant challenge again:...
That kind of stuff takes 1/10th as long to code as procedural decorations, and has way more replay value.
I think you drastically underestimate the difficulty of what can be summarized as "dramatically improve the Army AI of Hell". On the other hand, some of the stuff you're talking about will be coded in anyway as part of the Army Arc, such as better handling of the difference between melee and missile troops, better handling of building destroyers for sieges, etc. Having a routine capable of recognizing the need to sending a squad of trolls forward to destroy a gate while under the cover of goblin archers is not fundamentally different from sending a group of steel demons forward to destroy a door while under the cover of fire demons.
I've got a meta-objection to your construction, however. You seem to be taking a very gamist approach to Hell, which I would characterize as "Hell is the final boss of a game, and should be a more interesting and complex fight".
I think that continuing the developing procedural nature of DF will result in a more interesting and varied experience in the long run. We're already starting to see the Spheres play a stronger role; of particular note is the deities who allow demons into the world to serve their ends (e.g. a deity of war assisting a demon to raise a spire because it will cause more fighting, etc.). I think most demons should have a sphere, and they should act accordingly.
To give a few examples:
* Demons of combat, war, etc. should have full advantage of the Army Arc improvements, and have a focus on destroying your military units and equipment.
* Demons of gluttony should try and sneak or fight their way to your food and booze stockpiles, and consume them; basically a super-powered version of CURIOUSBEAST_EATER and CURIOUSBEAST_GUZZLER.
* Demons of filth should attempt to spread miasma and syndromes in your high-traffic areas and meeting zones.
* Demons of chaos should attempt to pull levers, open cages, toggle gears, and otherwise cause mayhem.
* Demons of corruption should attempt to spread transformative syndromes into your water supplies and stockpiles (think vampire blood in the well, but more along the lines of "roll on the random mutation table, and loose significant happiness").
* Demons of hubris should path to your highest-ranked noble's area, and attempt to reduce its value... destroy statues and furniture, deface engravings, etc.
* Demons of fire should want to see things burn... dwarves, doors, stockpiles full of bolts... path toward the largest nearby flammable thing that isn't yet burning, set it on fire, repeat.
And so on. The exact details don't matter; but the core idea is that you almost always get more interesting villains when they have a motive other than "be a punching bag for the hero(es)". IMO a sphere-based goal system for the demons would generate the most interesting stories and re-playability. For instance, one fort might end up facing primarily a brutally efficient military force, that ignores distractions and is bent on crushing your defenders, siege equipment, and traps; another might face mostly a whirlwind of chaotic and corruptive beings that burst past your military forces to pull levers, poison water supplies, consume your supplies, and turn your livestock and civilian dwarves into hideous abominations.