idea about mana

[Xecutor]

Let's say our wizard has 100 mana.
He casts a spell that cost 40 mana and has 'return delay' equal to 6 turns.
'return delay' is number of turns after which mana spent for spell returns to pool.
So after 6 turns wizard's mana pool will be increased by 40.
Buff spells will reserve part of mana pool which is returned when buff is turned off/dispelled.

[Bear]

That could work.  It seems like a sort of cross between cooldown periods and a classic "reserve/recovery" mechanic. 

There's an interesting sliding-scale of wizard power between "spells cost a lot of mana but you recover fast," which means you rarely worry about it except in combat when you have to manage your mana pool turn-to-turn, and can run out even in just a few turns,  and "spells cost not much mana but you recover slowly" which means you have plenty o' power for a fight when you're fresh, but have to manage your mana pool carefully, over game hours or even game days. 

Your twist here is that the recovery rate / mana cost tradeoff would depend on which spells you cast.  You could have something like a magic missile take one mana but tie it up for an hour, so you can cast lots of them fast in a tight spot but it's not a good long-term strategy because it will leave you exhausted and vulnerable, and something else that takes a lot of mana but ties it up for only a few turns, so you can do it once every ten rounds or so but have to work on "not being seen" during the rounds you can't do it.

[Hi]

is it better to have the mana return in a lump sum or is it better for it to return it continuously between now and then?

The cool thing about this is that you can have a spell to borrow mana from the future (checking first that you actually have that much mana in the future).

[kipar]

At first sight it seems a brilliant idea to me. But after that I've done some maths.

Ignoring side-effects (let's assume that all casts and all effects are instant) we will have something like this (see picture). It is easy to see that many short timed, but costly spells will be equal to many long-timed but cheap. But these two types of spells can intefere bad.

So, two conclusions:
1. "Cost" of the spell is proportional to (ManaCost)*(TurnsToRecharge)
2. In addition to managing his scalar mana resource between different spells player will be forced to play tetris-like minigame of optimal distribution rectangular areas in "mana-time" bar.

This can be fun, this can be boring. Everything depends on other game details and on the developer.
My own realization perhaps would be boring.

---
EDIT: removed words about "linear return".

[Xecutor]

Ignoring side-effects (let's assume that all casts and all effects are instant)

That's very brave assumption.
If you cast N spells in single turn, than, yes, it's as you drawn.
But spells are being casted in subsequent turns at the very best .
That's where short delay kicks in.
If wizard has 100 mana, and casts two spells with 50 mana cost and 2 turns delay,
he will regain full mana in 3 turns (1st spell on 3rd turn, 2nd on forth, after 3 turns).
If he casts spell with 10 mana cost and 10 turns delay, he will regain full mana in 10 turns.
While cost*rechargedelay is the same, impact on mana pool is very different.

[kipar]

Impact is different, but effect is basically same (if you don't mix these two types of spells):

Spell1: mana cost 10, recharge 10
Spell2: mana cost 50, recharge 2

1. If wizard use only cheap and slowly recharging spells, he will be able to use next spell right after the casting 10 of the spells in a row. His mana will be low (10-20), but it doesn't matter because his spells costs low.

2. If wizard use spells for 50 mana but with fast recharge, he will be able to cast spells at approx. same rate (not every turn though... perhaps right formula is cost*(rechargedelay+1)), but his mana will be full most of the time.

3. Tetris starts when you start to combine those spells. In your example wizard must use either 0spells1&repeat 2spells2, or 5spells1&repeat 1spell2, or 2spells1&0spells2 for optimal manapool use. I believe that non-instant spell casts makes formulas somewhat more complex, but the basic principle is the same.

[Xecutor]

Well, it's true that you can cast equal number of spell which have equal cost*reload delay values.
But that doesn't make them equivalent as long as combat do not degrate to standing at one place and repeatedly casting single spell.
At any given moment state of mana pool is different for spells with different cost, but equal cost*reload delay.
As in example above:
If you cast spell with mana cost 50, you cannot cast spells with cost 80 next turn.
But if you cast spell with cost 10, you cannot cast spell with cost 100 for next 10 turns!