That's a pretty interesting idea. As it is now, I don't think it's worth it, but if you take the idea bit further, you could have it so that you solve SB, but each edge can be in any of the three positions. Then, when it comes to CMLL, You do one of six algs to solve corners and SB at the same time. I might check that out some more.I'm not sure if this is useful, but I came up with a new way of doing SB + CMLL in Roux.
Imagine you have a cube with solved blocks. When you do R2 U2 R2 U2 R2, the second block will have what is basically an equator flip on Square-1. In this method, you will solve SB into that state, then do an algorithm so solve CMLL and flip the equator.
Solving SB is super easy, you simply make each pair with only one color matching in each pair, and insert the pair based on where the corner goes.
There are a few ways to solve CMLL + equator flip. The easiest is to solve CMLL, but cancel into R2 U2 R2 U2 R2 at the end. This works really well, but I've almost finished generating unique algorithms for this step. Most of them are garbage, but some of them are pretty good. Examples: R' U2 R2 U' R2 U2 R2 U R2 U2 R, R U' R' U R U' R D R D' R D R2 D' and R U2 R' U2 F2 D R D' R' F2 R2 U' R'. Most of the time canceling into R2 U2 R2 U2 R2 works better though.
Obviously this wouldn't be used every solve, as it would add on average 4 moves to each solve, but I think it's useful if you already have a pseudo-pair solved, or a free pair.
R2 B2 F2 D2 L2 U' R2 F D2 B L' B L2 R' D R' D2 B (FB solved)
U r' U' r U R U M U r U' R' //Psuedo-SB
R U2 R' U' R U' R U2 R2 U2 R2 //CMLL +equator flip