You are both sort-of correct. Mages grow near linearly up until they start hitting the maximum, and then they level off, taking better frame rate/better internet connection to achieve a difference at the absolute limit.
At 0.5/s spell, you are nowhere near this limit, so itās a true statement to say that you can double your output by casting 1/s. After 2/s, yeah, youāre probably reaching a limit and itāll taper off. Improvements in skill are limited by framerate, hardware, and the sheer capacity for the server to register things (along with human limits on speed). It does not plateau, but ends logarithmically at some theoretical asymptote that nobody could ever hope to calculate.
Other classes, however, plateau. See precision ranger which gets a theoretical (and calculatable) maximum damage for hitting precisely 1/s at a certain distance. Does it scale linearly on the way to the plateau? Not so much, because of the complexity of the way the timing and aiming work. For lower capability, a smaller change in skill will increase your damage tremendously. As you increase in skill, the amount you can increase in damage is reduced. For higher skill, though, you hit that cap. Basically a logarithmic curve at first with a hard cap, rather than an asymptote.
The same is true for musketeers, scoundrels, shaman, etc. There is a huge curve thatās hacked off at the top for an enforced plateau.
What this means is that mages have a theoretical limit higher than everyone else, because itās not truly capped but asymptotic. More importantly, it also means that unlike other classes where the initial stages of learning to play can raise your effectiveness substantially (logarithmic), you have a linear effect and so crappier mages are not as good as equivalently skilled rangers, scoundrels, etc.
The way to fix this is to make the curve logarithmic. If you give diminishing returns to faster casting, you have achieved this. What you also do, though, is lower the asymptote, which would not be good. So raise the asymptote by increasing base damage.
Then, to make every class balanced, move the other class plateaus to be in line with some % away from a theoretical limit of runemage (remember, canāt calculate the asymptote, but you could probably approximate it).
This methodology has the benefit of maintaining the importance of runemage skill (Cam and Sift and J get to be better than everyone else) whilst also making other people and class combinations viable at end game. Win win.
Edit: obviously do NOT enforce a hard cap on runemages. That would be a terrible way to fix this, and is not what Iām arguing for.