For Priests, all our mana is "base mana". For the sake of argument, let's use these numbers, using Heal for comparison.
In the new system, we have 100k "base" mana. 2% of that is 2k, so the Heal costs 2k mana.
In the old system, we have 20k base mana, which is increased beyond 100k by int. But Heal costs 20% base mana, so 9% of 20k is 1800.
So you can kinda see that the raw mana costs remain relatively the same. The big difference now is under the old system, spell mana cost as a percentage of total mana, got smaller and smaller as you got more intellect. Under the new system, mana cost as a percentage of total mana will remain the same all throughout, which means more spirit is needed to help your mana.
Might as well add it here. The reason priest mana costs listed on the front page look like they all went down, whereas all the other classes went up, is because for a priest, 100k mana is all base mana. For a holy paladin, on the other hand, of his total mana, 20% of that is his base mana. So a spell with 30% base mana cost would be 30% of 20% or 0.3 x 0.2 = 0.06 or 6% of his total mana.
This new system has some interesting possibilities.
Total mana available during a fight: starting mana + spirit-mp5 (assuming they do away with all forms of active mana return). Both starting mana and mp5 are linear with spirit. Say:
Total mana = base*K*spirit + L*sqrt-int*spirit (very vague general terms here, does anyone know the new spi-mp5 formula?)
Size of heals: same basic concept as before, modified by spellpower. For instance:
GHeal = base + spellpower*K
Cost = 2000 (as a guess, 2% base mana. I have no idea what it really will be, this is to illustrate something)
The point i'm trying to drive at: if you try to evaluate how total healing during a mana-intensive fight behaves, you wind up with spi in every term and int in only some. For cata, since spirit only affected a small percentage of your regen, lots of classes could neglect spirit, as its value only became evident on extremely long fights. In MoP if spirit gives starting mana AND mp5, its value will be much more consistent regardless of fight length.
Total Healing: Heal Size * (Total Mana / Heal Cost) = (Base GH + SP*K) * (Base Mana*L*spi + M*sqrt-int*spirit)
very simplified, TH = [base GH*K + base Mana*M*SP]*spi + [base GH*L + SP*N]*sqrt-int*spi
Notice: all 4 terms have Spirit in them.
As for int (sp = 2*int):
one term doesn't even have int
one term only gets sqrt(int)
one term has int
4th term has int^3/2
IF the spi-mp5 formula changes to omit int entirely, total healing will be even less dependent on int than I'm suggesting (2 terms will omit int, two will have int multiplied once).
Why in the heck does any of that matter? Well, it's a question of trends. Basically spirit will be an overall total healing stat and should actually produce much better results than int in that regard (if they do the scaling correctly, and if we're never allowed to reach a regen cap). Int will become a burst stat, like haste. It'll give us higher total healing too, but not nearly as much as spirit. Even on very short fights.
As always, who knows. Maybe i'm completely off my rocker.
To make this model work, I think (1) tune regen so we're struggling with OOM the whole expansion (2) make our base heals (Heal, Spirit Shell) far more efficient than they currently are. At a very minimum, both need to be significantly more mana efficient than GHeal.