My two cents on tactical combat:
In the video, FB's attack on the slowed enemy could actually have be his best option. The reason is that the un-slowed bandit (or whatever it is) had just attacked, hence FB decided to attack the opponent that was next in the initiative bar. This is the best tactic if you have a good chance to kill your opponent with a single attack (which indeed happened) since you could kill them all before any of them has a chance to attack again. If there was a mistake in his strategy, that was probably to cast slow on the opponent that was not ahead in the initiative bar.
About the discussed STALL when the AI plays smart, that would happen only if the units have the same movement. Otherwise, once the higher-movement unit gets you in its range, you are still too far to catch, so you'll end up being attacked first (the next turn if oyu stay where you are, or some turn later if you keep retreating). Despite small details, I am all in for an AI that plays smarter, even if not smartest, hence I think Seanw3's and HF's concerns after watching the video were legitimate, though hopefully unnecessary 
As for terrain in tactical combat, I'd ask you not to include it, since this would likely make the AI worse off (too many things to account for). Superb games like HoMM3 and 5, or MoM did not account for terrain - expect for impassable terrain or minor things - and they were a lot of fun. HoMM4, which I have not played, accounted for line of sight, and I heard that it was not as good as the others. So bottom line, sometimes less is more.
On a side note: I have not played the game as yet, but I do hope that ranged units get some appropriate malus when shooting too far, or too close, or to single targets, or when disengaging from melee units. In general, I'd say that an eccessive strength of ranged units is one of those mistakes that can make tactical battles boring and remove most strategic considerations.