Here is some of my Feedback. Its late here in germany, so i will only touch those aspects that come first to my mind:
Caravans. The income from Caravans once you have some cities is quite ridiculous. This is caused by the number of caravans growing quadratically with the numbers of cities [2 * (N-1)^2]. Removing the ability to send a caravan from A -> B AND simultaneously one from B->A will slightly reduce the caravan impact, a limit of one caravan per city (that might be increased by diplomacy research) and should be capped (in endgame) somewhere around 3 per city will balance caravans i think.
The other aspects of Diplomacy are still a bit obscure. The availability of Diplomatic capital only throuth the embassy seems a bit limiting here. Some treaty opitons are not really available in 3C.
Dependence on Metal: Not given, as Gildars are required in much higher amount. Therefore, the real army production (neglecting the upkeep) is more dependent on gildars than metal.
Right now the weapons are not differnt enough, this might of course change when special abilities are introduced and hopefully awarded, in part to the weapon of choice. While others oppose the idea to further divide weapons according to their strength I feel that technology should not provide within some turns as strong jumps as 3 -> 5 -> 8 -> 18 -> 30 (available arounf turn 30). The maximum weapon damage should be chosen lower and the damage of the starting weapons should be higher, so more along the lines 6 -> 9 -> 12 -> 15 -> 18. This would allow meaninful soldiers even at the beginning.
The cost balance of Defense vs. Offense is strongly screwed towards Offense.
a) There are no cheap defense tools available.
b ) Good defense is by a factor of 10 more expensive than compareable offense (while realistic this is a balancing problem) and
c) there is no need for defence when you can kill the enemy before he can even retailiate (which is possible due to the strong offense).
a) In Tactical Combat calculate retailiation simultaneusly with the attack. Defense is then also valuable to the attacker.
b ) Put a lot of the defense capabilities on shields and provide a shield that is initially available and provides a defense that is at least half of the inital weapon strength (about 4 or more).
c) Increase Health points of all units and heroes
One problem here is, that you willl not get additional heroes if you switch late to adventuring mode. Some heroes should spawn only for adventurers and should only be hireable by them (also higehr level champions hsould be available). Furthermore, at the moment adventuring is not viable on small maps, as there are simply not enough quest locations (and some of your goody huts are stolen by npc adventureres). This could be remedied by continously spawning new quests. Another disadvantage of the Adventuring pathway might be, that Heroes most probably will simply be overwhelmed by larger Army contingents later in the game. At the moment they seem quite powerfull, but this is only caused by the fact that they can kill nearly everything without fear of retailiation. Units within AI armies are usually killed with the first hit and therefore no damage is taken, strong monsters can be softened with magic well enough, to take them out without damage. This is caused by the high tactical combat speed and the fact that at the moment the player typically starts the tactical battles.
The problems with the available quest types have already been mentioned.
A lack in magic is the availability of stonger damage casts. In the beginning of the game doing INT dmg seems also quite powerfull, hoever when confronted with real armies a caster uses up all his mana during the first turn of combat. Afterwards he is quite powerless, and that is for about twenty turns (on the main map) until the mana has come back. Therefore the best way to play with magic at the moment is to use the spells from the book of mastery and eleminate enemy armies and cities without entering tactical combat (which is also fun and reminds me of slpha centauris planet busters). The problems with magic are at the moment still not too obvious, because casters do not use up mana when auto resolve is used.
Overland Magic and the Feedback given by the Game
Furthermore the magic part of the game needs some more spice. Of the large amount of spells on the main map only raise and lower land and the spells of the mastery book really stand out, because tehre are visible results. The impact of the other spells is not visible enough, so most often one does not bother to use them at all.
I will give one example: When you clicked on a enchanted unit in master of magic you saw the buff spells that were casted by you and you saw their direct effect upon the units strength (the swords on yellow background). You furthermore saw a aura in the color of the element of the spell around your unit. Therefore the game gave direct feedback what your spell did. In elemental such information is not visible at all (furthermore the power gain of units through magical buffs is rather low, and the number is low as well).
the limit of summons to one per type does not allow to use an entirely summoned army as executive force (ok that is also caused by the bad placement of caster souvereings in front of their summoned minons in tactical battles). Therefore the caster has to perform the things himself. With the current balancing the souvereign is either dead, before his minons did the task, or has killed the enemies himself.
Civilization is the most research oriented tree but unfortunately you cannot research a lot of things within the tree right now. The food and gildars come from caravans, and the other resources are not used in meaningfull amounts. Therefore the civilization aspect is a bit underwhelming at the moment. This might get more important through when caravans are balanced, and the cost of Warfare is moved more towards resources instead of gildar. Also the items buyable for your heroes could require resources instead of gildar (at least in part).
More better thought out feedback will follow after some sleep.