An example like that doesn’t really fix SMN’s problems because SMN’s problems are the fact that the rotation is far far too rigid and each primal doesn’t do enough to justify its own existence
Let’s say we apply you idea in practice. S/L/R have 2 long cast primals between them and one of them regens mana. There is no choice. The primals are designed to be exactly long enough to fill the window between your Demi’s so shiva’s mana regen is just a free given. With 2 hard cast primals pheonix is just arbitrarily less mobile than Bahamut; but you can’t control which Demi you summon next so you can’t play around that, you just have to be less mobile during that phase with no recourse. Compare that to something like BLM with triple cast or PIC with weapon motif.
As long as you can’t control which Demi you summon next the primals amount to visual flavour you can’t do anything with. However if you introduce choice then one dominates and leaves the other one in the dust, let’s use your example again. Let’s say both halves do the same damage and you retain SMN’s current rough mana neutrality. You’d never use S/L/R because T/G/I is more mobile. Let’s say that SMN was made strongly mana negative, then T/G/I would barely be used. If T/G/I did more damage but SMN was strongly mana negative. That could potentially introduce a miasma style 2 choice
Regardless all this hinges on choice being introduced, the current rigid rotation with S/L/R really adds nothing to the job. Adding internal complexity to the current 4 phase rotation per minute I think is a better way forward, I’d even be in support of making them longer and more complex but adding limits, like each phase is 35 seconds but Bahamut is still on a 1 minute timer but you need to cast all three astral flows to proc Bahamut to be able to summon. So how much of each phase do you skip by overwriting it with a different primal to maintain the 1 minute loop