Frogboy et al,
Longtime listener/reader, infrequent poster here. Lots of ideas (and rants) being thrown about. I like some, don't care about a few, and think most depend entirely on your implementation.
I'm actually really curious why you've thrown out the real time aspect of the tactical battles. I can't say that I have a strong preference here, but the real time aspect definitely made me consider the game more unique. Something about the implementation must have been causing battles to be un-fun? or was there programmatic problems in achieving what you envisioned? At any rate, just curious.
Combat speed - This is one that completely depends on your implementation IMO. On surface, I agree with some other posters that think this should be broken out. Something like initiative (which determines who attacks 1st, 2nd, 3rd etc.) and action points (where attack, defend, cast. etc. take up 'x' # to perform). Sounds like there is some conflating this concept with area of attack weapons/spells judging by your Sauron example. Thought should be given to make these distinct as much as possible I feel.
Morale - Again devil is in the details. I like your initial thoughts and some ideas tossed about, but will wait before recommending tweaks.
Terrain Randomization v. Richness - Well, I really liked the idea where these were combined and tactical maps get randomly generated for the most part while there are also some key points of interest (POI) with pre-constructed tactical maps. If this is straight-out not feasible, then I would hope some sort of compromise of these two would be possible (e.g. Mostly pre-constructed but considers major territory on main map). If internal resources/cost are a limiting factor in implementing this design, and its an either or choice, then I would vote for randomly generated.
WTA - Lots of concepts to deal with here. I'll strive for brevity and clarity.
1) I don't really like an arbitrary end to the battle in a given number of turns 'N'. Perhaps, you could implement a feature where you ask the attacker at the start of battle what type of attack they will be conducting (siege, guerrilla, regular army clash, etc.) and this determines how morale, retreating, losses, etc. are determined. Or you can ask how many turns 'N' the attacker thinks is required for their attack. Anybody looking to run a guerrilla attack enters 1 or 2 and then retreats with minimal effect to their army, siege takes long time (might be determined by supply lines), etc. Just some thoughts.
2) I think it makes more sense to allow retreating in some fashion, but perhaps penalize the attacker/defender differently. An attacker wishing to conduct guerrilla raids won't see a massive loss or his army separated at the end of battle, while a defender (if they can retreat at all) incurs heavy losses and an army dispersed at battle's end.
3) Definitely don't like the construct where heroes always escape. Don't like it either that losing a battle means your hero doesn't have a chance to withdraw (unless in a siege situation). Implementation will make or break this concept me thinks.
Combined Arms - like it.
Thresholds - like the direction, will wait for implementation. I really like the concept that you can fight some or most of a tactical battle then hit auto-resolve at any time. This lets you carry out a few key decisions like 'cast fireball' or 'mass heal' then let the computer handle the grunt moves.
Battle Length - Not sure how this maps to a TB tactical battle. Time limits make more sense in the paradigm of Real time or multi-player battles. Since you've moved past real time, perhaps you can have players set this "battle length" at the start of a multi-player game, or set tactical battle turn time limit. If we're talking about number of turns, See WTA #1 for my only real value add idea on this.