As I said in a previous thread, setting a limit is reasonable, but setting that limit to one is not reasonable. I see no flaw or fault in allowing n units per tile, if n is a reasonable compromise.
There is no possible compromise.
Any stack size of size n (where n>1) reduces your frontage (where frontage is defined as width of the minimum {tile space} your soldiers require to fight effectively) by (frontage/n).
It should be obvious that the greatest absolute drop in frontage size is the change from n=1 to n=2.
Frontage and reaction costs (where reaction cost is defined as the time and expensve of moving a unit from one extreme of a formation to the other, and then responding to the attack) are directly proportional -- as one rises, so does the other. Both frontage and reaction costs are inversely proportional to concentration (where concentration is defined as the number of units able to fit in a single tile).
By definition, the defender gains from lower reaction costs -- the attacker has already concentrated and committed their troops for the attack, and the defender must react by concentrating their troops after the fact. As reaction costs plummet and concentration rises, the defender becomes stronger and the attacker weaker.
cIV is a textbook case of the effects of an unbounded positive n. Defense was a trivial task for a force roughly equivalent to the enemy ( If(location == forest hill) {defender wins}), while offense was something that the AI was, practically speaking, incapable of even in AI on AI wars. They could win offensively if they overwhelmed the defender through tech or numbers or both, but not against an equally matched foe.
The player, on the other hand, was able to offensively dominate through strategic concentration (stacks on the border before wars start), rushing and a refusal to engage in offensive action on a target not currently occupying a city unless the odds of winning were better than 80% -- thereby forcing the AI to eat the costs of field warfare.
While n=2 is not nearly so bad as a positive unbounded n, it is still significantly worse than n=1 for the reasons described above. Which is why, as I said, compromise on this subject is impossible.
tl;dr
The player should not have a monopoly on offensive action. n>1 reinforces that monopoly (see above for why). Therefore n>1 is bad.