There is a whole sequence of increasing redundancy.
- One tank, one 1st stage, one 2nd stage.
- One tank, one 1st stage, two 2nd stages.
- One tank, two 1st stages, two second stages
- Two tanks, manifolded, one 1st stage, one 2nd stage. (No one does this any more).
- Two tanks, not manifolded, each with one 1st stage, one 2nd stage.
- Same, but manifolded.
(1) was "the way" for many years.
(2) is now "the way," for the last 40 years
(3) suggested for cavern, ice-diving
(5) common in sidemount, no longer common in backmount
(6) standard backmount configuration
@Angelo Farina suggests (3) is the minimum needed.
My current tank (15 liters steel, Faber) is number 3 PLUS RESERVE, which adds another significant safety, trapping 1/4 of the tank capacity, which is preserved in case of a sudden air loss (O-ring extrusion, for example), or in case the diver forgets to check his SPG.
Please note that I am a purely recreational diver and instructor, I did never practice anything requiring a technical certification.
However here in the Mediterranean the limits for recreational diving are a bit higher than in other places, as here we consider recreational diving going down to 50m and with deco on back gas (not accelerated deco with highly oxygenated mixtures, this is technical also here).
So you can understand how for deco diving in the sea having some air reserve is prudent.
If your tank does not have a reserve, an additional pony tank or a true twin tank is needed for these dives.
It must also be said that I am growing old, so I do not have plans to make deep deco dives in air anymore.
As one of the two valves of my tank is DIN-capable (unscrewing the adaptor), I have now configured one of my regs (a SP MK10) with the DIN conversion kit and two identical second stages (R109 modified to BA).
My plan is to use it also on rented tanks for low-depth dives, during the next summer, where objectively two independent firs stages could be excessive.
With yoke regs, instead, I would never use only one first stage: I have seen too many O-ring extrusions (most on the boat, but several also underwater), and they are not nice. My son also experienced a completely clogged first stage due to rust from a rented tank.
So it is quite evident that case 3 is better and safer than case 2, without the added complexity required by two tanks.
Regarding the typical American approach, that is case 2 plus a pony tank, I consider it more bulky, more complex (in the end you have THREE second stages) and not significantly more safe than case 3 with reserve. There are particular cases when also I did use a pony tank, but it always occurred starting from a manifolded twin tank with two regs, when planning for deep and long dives, or when requiring redundancy for other reasons.
Last remark. The list is not complete, because it does consider the presence or absence of the reserve (which in my opinion increases significantly the safety beyond the corresponding case without reserve).
And, in case of manifolded twin tanks, it does not separate the case of a permanently open manifold (as most "old" twin tanks were, including my ones) from the modern manifold with a separation valve (which is considered an alternative to the reserve, albeit, if I can choose, I prefer a twin tank with a permanently open manifold, two valves and reserve, to the modern configuration of a separation manifold and no reserve).
