If a patent newbie might be permitted to comment, the problem seems to be people falling into two general camps of thought which are diametrically opposed. To wit, when the Sovereign is defeated in battle, either he dies and, literally or effectively, ends the game, or he is somehow removed to a location of relative safety, permitting the game to continue, albeit with the defeated suffering some disadvantage. They both have their appeal, the former making gameplay more decisive (although it might often proceed more slowly due to players being more cautious with their Sovereigns) while the latter gives more opportunities for upsets and reversals of fortune. Neither approach can be said to be definitively superior than the other, so people are trying to find some way to blend the two ideas.
Now, the effects of a Sovereign's defeat will greatly impact how players must operate, but the effects themselves will be a rather small portion of the game that could easily be adjusted at need. That in mind, the best solution, it seems to me, is no solution. Simply allow an option, in setting up a game, to choose which model the players want to use. The specific details of the perma-death and respawn models could also be available for adjustment.
Details of the perma-death model are fairly minimal. If the game ends immediately, that's that. I would suggest in rounds with more than two sides, however, that, so long as at least two Sovereigns remain, sides having lost their Sovereigns remain functional for a time. Those sides could start collapsing inwards as the leadership and direction structure provided by the Sovereign are gone, but they should still be able to influence the course of the game, whether it be by launching suicide attacks to weaken a particular side or offering their remaining land, forces, and resources to be absorbed by another Sovereign's side. The powerful conqueror could find himself abruptly outmatched by an opponent having absorbed the remains of the conqueror's previous victories. This brings in diplomacy as well, since a side devoided of its Sovereign would, reasonably, be more likely to allow itself to be absorbed by a friendly side.
As for wanting a less risky use of the Sovereign, two parts have to be considered: First, what are the exact mechanics? Second, and almost as important, what is the story cover for those mechanics? Of course, the issues may be considered in reverse order, which is the route I took. Being sent somewhere, in some condition, is obviously a requirement to the respawn model, so some sort of story must be concocted to cover it. The idea that a grievously wounded Sovereign could slip away moments before death and get anywhere, much less back to his capital, seems flimsy at best. The consensus seems to be that essence should be lost by defeat, which isn't particularly well explained by the above, either. You could say that the serious wounds cause the Sovereign to lose his metaphysical grasp on part of his essence, but even that feels contrived if winning a battle in which the Sovereign is nearly defeated doesn't cause the same penalty. My version is based on the idea of essence-powered spells, the first and most basic being something akin to Word of Recall. This emergency spell would be precast, binding a certain amount of the Sovereign's essence for use when the spell needed to be triggered. The spell could also contain optional modifications, such as upgrading restoring health, changing the target of the teleportation, pulling friendly units (remaining in the battle or from elsewhere) with the Sovereign, causing damage to enemy units in the battle that defeated the Sovereign, etc. Each option could adjust the amount of essence required for the spell, and the Sovereign could even choose to refrain from using it (the pre-game option to allow respawning could be phrased as "Allow [spell name]:" with a check-box). The essence used for the spell would, of course, be consumed upon triggering the spell, but both the manner in which required essence is determined and exactly what happens to it could be adjustable. For instance, the essence could be removed from the Sovereign's pool functionally, while still impacting things such as the mana cap (assuming it is still tied to essence), being as the essence exists, but simply isn't available for other usage. The amount of essence needed could be determined as a percentage of the Sovereign's total (feasible on the basis of requiring more power to move a more powerful Sovereign), a static amount, or even a combination of both (a percentage for options directly related to moving the Sovereign, and static amounts for things like healing and damage dealing). Last, the used essence could be either permenatly lost or slowly returned over time.
Personally, I would like to see essence completely delinked from the mana cap and treated as a separate resource and as many pre- and in-game options as possible for adjusting the effects of a Sovereign's defeat. Here is the system as I imagine it (all specific phrases, wording, and numbers are just off the top of my head):
Pre-game options:
Check-box for "Allow Emergency Recall [or whatever it might be called]:". (self-explanatory)
"Essence requirement:" and options in the vein of high, medium, or low. (adjusts the amount of essence required for the spell and its options to make their use more or less risky; percentages and flat amounts are both adjusted)
"Permanent essence loss:" and a slider from 0 to 100. (adjusts the percentage of essence used by the spell that will return over time)
"Essence recovery rate:" and slider of high/med/low options. (adjusts how fast the temporarily lost essence returns)
Check-boxes for allowing the various in-game options in the spell. (detailed in the "In-Game options" below)
"Casting frequency:" and a slider from 0 through some arbitrary number. (adjust how often, in turns, the spell can be cast, 0 being unlimited; could be used to decrease the grind of repeatedly crushing an inferior foe)
Allow modifying (lets Sovereigns recast the spell to change its options)
Death options:
Essence dispersal (what happens to the Sovereign's remaining essence when he is killed)
Dissipates (poof, gone)
Distributed evenly (all remaining Sovereigns get an even share)
Distributed by diplomatic relationship (remaining Sovereigns get shares allotted by relationship with the deceased; requires a diplomacy system with enumerated or enumerable relationships)
All to most friendly Sovereign (self-explanatory)
Check-box for "Retain control after death:"
"Ignored orders frequency:" and slider for percentage (commands of all types have a chance to be ignored sing the orders are now emanating from the land's non-Sovereign second-in-command, who is having a hard time maintaining control of a disintegrating side)
"Collapse level:" and slider or high/med/low options (towns farther from the capital have higher chances of defecting to other sides or just going neutral; this is a universal modifier to all towns' chances to break away; indirectly modifies how long a Sovereign-less side can exist, as a lone capital will mutiny, ending the side)
"Max defections per turn:" and slider (self-explanatory)
Check-box for "Allow ceding of towns:" (allows side to give towns to another side that has a Sovereign)
Likewise for forces and resources
In-Game options for the return spell: (essence cost a basic amount pegged to the Sovereign's level plus whatever is required for selected options; essence used for the return spell is simply held in reserve, meaning it still counts toward the Sovereign's total essence but is not available for any other uses; whenever a required essence amount is updated after the return spell is cast, like the Sovereign leveling up, the required addition is removed from the Sovereign's pool and a notice to that effect is given; insufficient essence to cover the change will collapse the return spell, requiring it to be recast)
Return to selected city (sends defeated Sovereign to the selected city instead of to the capital, closest city, or wherever the default is; percentage essence cost)
Healing (could be handled lots of ways: selectable flat amounts, selectable percentages, fixed percentage, amounts or percentages pegged to level, etc.; essence cost depends on those)
Pull surviving units with Sovereign (teleports remaining friendly units, strongest first, in battle along with the Sovereign; selectable flat amount of units to bring; essence cost a flat amount pegged to the number of units allowed for)
Pull unit to Sovereign (selects a unit to be teleported to the Sovereign's upon his emergency return; can be added multiple times for different units; essence cost a flat amount pegged to the unit's stats in some way)
Parting shot (two versions, probably just have one available; essence cost could be all over the place)
Direct damage (simply deals out damage to remaining enemies when the Sovereign is defeated; essence cost based on compounding costs from sub-options)
Number of targets (maximum number of enemies that can be hit; could simply be all or damage spread evenly)
"Target strongest", "Target most valuable", or "Target lowest life" (assuming the parting shot can't hit all enemies in the battle, chooses targeting priority based on most powerful, most expensive to produce, and most likely to be killed, respectively)
Size of blast (range from fallen Sovereign to damage enemies; this is an alternative to the "Number of targets" above)
Elemental nature (determines the type of damage)
Piercing (damage ignores armour/resistance)
Spell (casts a spell on triggering the return; could be added multiple times for different spells; could be offensive spells or spells that affect the enemy army outside of battle; essence cost based on the selected spell and power)
Spell power (assuming spells have scalable powers)
Target (could use the same options as for "Number of targets" above with addition of "Self" for the point of the Sovereign's defeat)
Mana boost (provides a selectable amount of mana upon triggering the return spell; essence cost a flat amount pegged to the amount of mana selected)
Chaos rift (cracks reality during the teleportation, creating a random group of monsters, random damage or healing, random environment changes, etc. at both the from and to locations; reduces essence cost of the return spell by a percentage due to less effort required for the sloppy teleportation)
Directed rift (the chaos rift only affects the from location; changes the essence cost to a positive amount and doubles it)
Astral assistance (summons random creature to your service; essence cost dependent on sub-options)
Power (determines general power level of creature that can be summoned)
Duration (how many turns the summoned creature will remain)
I'm pretty sure I've forgotten things here and there, but that's the general idea. Giving players the option to set up the way defeats are handled wouldn't be exceptionally difficult or even significantly alter the game, but I imagine it would go a long way towards pleasing a larger audience. Allowing options in the effect of the return spell adds another layer of tactics that could become vital towards the end of the game. I can imagine a seemingly-weak Sovereign being defeated only to march out from a nearby town, mostly healed and with several of his strongest units, to hunt down an enemy army badly damaged by the battle and a crushing parting shot. Anyway, that's my two cents'... *glances back up* er, five cents' worth.