I know I'm new here, but balancing issues, especially ranged/melee are ubiquitous. Regardless, take this as a "for what it's worth".
IRL ranged units exist on the battlefield (as opposed to only existing behind fortifications) because they ARE (or can be) devastating to other units, and because melee units DO take time to reach them. That's just reality. If that is well simulated in a game, then that is not, in itself, a problem.
As suggested in the OP, units with shields (e.g hoplites), especially if they are in good combat order, are generally well-defended against ranged units when not in melee. Therefor, a normal order of battle was to bring archers to the fore as the armies were arranging themselves, and then move them to the flanks where quick, and less well armored, troops would be attempting to flank their opponent's center (don't waste troops in ranged v ranged fights).
Setting aside the protection of shields, a huge reason armies weren't comprised predominantly of ranged units is because they can't shoot into melee without a generally 50/50 chance of hitting the friendly unit. That can be simulated by an in_melee flag which makes the unit no longer attackable by enemy ranged units. That would, of course, increase the cycles needed to find a viable target, but that might be balanced by losing a shot (or shots) as a shooter hunts for a new target after a set in_melee flag has been encountered. Another way to reduce cycles is for the shooters to search for targets by range. Check first the closest few enemies to shoot and then, if they are all in melee, look for the closest few at mid-range, and then for the closest at long range.
Of course the quickest and dirtiest method is to simply reduce ranged damage done to any unit while it is in_melee.
If you wish to be able to have the option of ordering ranged units to attack in melee, perhaps even limited only to large units like elephants and siege, then something like an in_melee_with list can be used to divide the probability of hit among the target and all units in melee with it, weighted by the size of the units. Whether that level of detail is feasible for a real time game of this complexity is, of course, another question, but limiting in_melee ranged units to very large targets may be doable, even if the best relative damage approximation is to reduce ranged damage done to those large units while they are in_melee.
