Actually, I'd even include production and trade between worlds as an influencing factor in such a game, as well as combat. I don't think it 'can't' be done, after all, Homeworld has been done, and so has MoO etc, so the two facilities are there.
I think the micromanaging was taken overboard in MoO and IG though, most leaders don't set, for example, the amount of money for a particular research, that's why they hire Cabinet ministers, who hire advisers, who hire consultant scientists etc, so whilst you could choose the general direction of research, you could allow race characteristics to define what they actually discover, it means a massive tech tree but you would only travel down 'twigs', some races simply will not think up certain ideas (the rules change for captured stuff).
As for fleet combat, theres no real reason why, using graphics at homeworld level, it could not have supported much larger fleets on modern computers, indeed, Nexus:The Jupiter Incident was actually the graphics engine for exactly that, for Imperium Galactica. So theres no real reason why combat could not be represented in such a way, without resource gathering, but with asteroids and other navigational elements. You'd have to vary the game pace on the events at hand, if you want high levels of fleet-strategy, then you need micromanagement at the cost of combat content, what I'm thinking of though limits the players control over a planets infrastructure, but still allows enough freedom to choose the direction of play, the content of your fleets and the direction of research.
It also means that new modded weapons and guns can be simply 'glued' onto the current tech trees, so only certain race-types could research certain hulls etc.
Yes, I've thought about this a lot 
My main concen is not with hardware or micromanagement issues.
I don't say it can't be done - it can't be done how RTS-es are made nowadays.
My main gripe is the lack of a capable Lieutenant AI and a Admiral/General level interface.
The later is even more important - it's time we forgot 'simple' point and click.
Whay your job should consist of is assigning objectives than handing them out to different battlegroups to achieve them. The actual gameplay part would be the variations on said objectives.
Right now only two objectives exist: move/advance/flee in terror & attack/capture/annihilate.
Nothing complicated, nothing that would grant you the flexibility needed for a true strategy, you're bogged down with tactical difficulties and tactics are nothing more than exploiting certain map features and herding your troops to said feature like sheep and keeping them from going off on their own.
In a true RTS I don't want to deal with that. If I gave an objective to my colonel - take the ridge North of town Vandero, purge out resistance, dig in and set up a perimeter than that's all I should have to do.
The AI would first conduct scouting on its own, asses the enemy's troops and installations - crossreference it with earlier Intel reports and send me a report of their own they differ too much, and warn me if they believe the objective far fetched (too much predicted losses, concerns over resuplying and/or lack of supply lines to evacuate) - then it would devise its own objectives (attack these troops, harrass those, sneak there....ect.) than hands said objectives down to his Lieutenants...and so on until it's just a grunt with the objective of taking that damn bridge.
The 'game' on the players part comes from the variety of objectives he can assign - harrasment, scouting, sneak attack, fast deployment, entrenchment ect. ect.
Beside you'd have the now 'standard' setting for engagement tactics to *refine* said objectives.
When commanding smaller troops and/or doing tactical command I'd like to furbish my own battlemethods before the actual battle so I don't have to micro-manage juggle all my troops in the heat of the battle.
1) Creating battleplans.
Forget the damn fog of war! Once the area is known - or I have a map for heaven's sake I should no longer have to deal with the treachearous terrain hiding 'suprises'. It should be the enemy who suprises me with having hided that long in the mud.
Give me intelligence reports for Christ's sake! Telling me the NMI is hiding out there is a big help. Telling me that armor collums were seen marching through Braska and there are infantry entreched at the river is a lot more specific. Adding stuff like telling that said unit is likely the 17th Gravediggers known for guerilla tactics is icing on the cake, but you get the drift.
Given all those data, you don't start point-n-click managing the battle.
First you set up the objectives. Then you're given the ability to determine how the commanders react if something pops up - ergo set up the next objective if they suceed, the retreat plan if they don't.
Radio rules - you shouldn't see *all* your troops *all of the time*, only the ones directly under your control with direct communication.
So actually before the battle start I should be able to drop objectives onto the map, then write scripts for my commanders.
Beside those reactive scripts I should also have the ability to define triggers - so instead reassigning a new objective to everyone once I have drawn the enemy out, I can merly yell - "Barbeating!" - in the radio and my troops start fleeing toward my core forces who were in hiding so far.
2) Training
I don't want training to be mere stats - I want to actually be able to preset/create tactics for my forces.
This could be preset battleplans the commanders use on their own.
Formation setup, breakup; covering fire; fire zone allocation; retreat priorities; regroup priorities and methods all fall under this category.
