I like your video. This is called ZBLS, our Wiki contains links to 3 algorithm lists if you scroll down. I wouldn't learn the full subset unless you already know ZBLL, but it can be useful to get yourself out of tricky cases.

A neglected reason for CFOP's dominance is that it can take a decade to become world-class. Even if Roux were to spike in popularity and become the most popular method today the top solvers would still be almost entirely CFOP, simply due to them having 10 years of practice and experience behind...

Just a thought - if you're using longer PBL algorithms to preserve the E layer you could solve the E layer before PBL. Perhaps while you're solving the first face, I tried some practice solves and it's relatively easy. I just don't see the benefit of using longer PBL algorithms to preserve the...

Not counting algs for the first face as it is easy enough to solve intuitively. The total number of cases is 35, 6 PBL cases multiplied by 6 equator cases and excluding the solved case.
The 6 PBL cases are the same as those used in Ortega for the 2x2, plus the case for all corners permuted...

Thank you for mentioning, but I feel I should point out the Game of Life and Death is a mobile game created a couple of years ago - the Game of Life is 50 years old

It depends on what algorithm you use. There's a correct and incorrect AUF you can use for the Z perm. Here's an example:
https://alg.cubing.net/?setup=M2_U_M2_U_M-_U2_M2_U2_M-&alg=M2_U_M2_U_M-_U2_M2_U2_M- (good)...

R2' U R //P3
U2 R U' R' //P4
U2 R U2 R2 U' R2 U' R2 U2 R //LL
Next: R2 U2 R' U' R' U2 R U' R U R2 U R' U R' U R' U2 R' U' R2 U2 R' U2 R'
Edit - was 2 minutes late, use Pyjam's scramble lol

This actually doesn't work, it's (R2 U2)5 R2 U which is 12 turns. However the optimal is 8 turns:
Depth 8
U L2 R2 D2 U2 L2 R2 D2
U B2 F2 D2 U2 B2 F2 D2
D2 U L2 R2 D2 U2 L2 R2
D2 U B2 F2 D2 U2 B2 F2
D2 L2 R2 D2 U2 L2 R2 U
D2 B2 F2 D2 U2 B2 F2 U
L2 R2 D2 U2 L2 R2 D2 U
B2 F2 D2 U2...

ksolve gives the following optimal solutions for the scramble U:
Depth 15
D L R D U B F D U L R D U B F
D L R B F L R D U L R B F L R
D B F D U L R D U B F D U L R
D B F L R B F D U B F L R B F
L R D U B F D U L R D U B F D
L R B F L R D U L R B F L R D
B F D U L R D U B F D U L R...

ELL is easier so learn those first. SV is probably worth learning, OLLCP is not. There's simply too many algorithms and if you know CFCE the time saved by OLLCP is negligible.
If you know WV and SV I would suggest an obscure LL algset - the cases with corners oriented. PLL and anti-PLL combined...

I really doubt it! Even among sub-4 times, a sub-3.5 is vanishingly rare. If I were to make a prediction I would only feel safe saying "we will get another sub-4 in the next year".

3-style is simply too move-heavy, as others have already said. Here's Jack Cai, one of the best BLD solvers, getting an 11.95 average with 3-style:
The only feasible way to learn 5-style is to find a method to create algorithms on the spot. You can't learn that many algorithms.

x2 R D2 //almost cross
R L' U2 L //P1
R2 u R2' u' R U' R //P2
U L U L' //P3
y M' U' M //finish cross
U R y R U2 R' U' R U R' U F2 U' //block+some setup
R U R' F' R U R' U' R' F R2 U' R' U' F U' //LL
FreeFOP start ending in Tripod finish
Next: B2 U B2 R2 U' L2 D F2 R2 U2 B2 U L' B F2 R2 U B2 D'...

y2 L F' R2 U D' R F2 //xcross
R' U2 R L' U L //P2
y R' U' R //P3
U' R U R' //P4
R' U' F' U F R U2 R U2 R' U' R U' R' //LL
CFOP, 34 HTM lol
Next: D2 F L2 U' B2 D' L2 D' L2 B2 D2 F2 U' R2 B' L F R2 B' D' U2

z U' R L' D2 R' F' D R //xcross
E' R U R' D' //P2
R' F U F' R //P3
U2 R' U' R //P4
U R U R' U' R' F R F' //C
M' U' M U M' U M U M' U M U2 M' U' M //E
CFCE
Next: U B2 L2 R2 D B2 R2 D2 L2 U R' F2 L B' L B2 L B L' D'