6x6 Last Two Edges
Cases named by what they do to the four wing slots of an edge, not by what the alg databases call them. Each alg below was simulated on a 6x6 so the description matches the actual effect; the 3D player starts in the case and plays the solve.
1Solve pairs
Get matching wings next to each other. Slice a wing's layer so it sits beside its partner on the other edge, flip that edge out of the way, slice back. The 3R slice pairs inner wings, 2R pairs outer wings, 2-3Rw moves both. No set alg; the shape is [slice] (R U R' F R' F' R) [slice'].
ABAB → BBAA
Alternating wings become two pairs. Done by hand with slice, flip, slice back.
2Align pairs
Turn the two edges until the pairs sit over each other in a non-matching way (AA over BB), so step 3's swap finishes both edges. If one pair looks "flipped" relative to its partner, that is usually an alignment problem, not a parity: realign before reaching for a flip alg.
Stack them opposite
3Swap between edges
Each alg exchanges wings between the front and back edges on one side (the side the player shows) and nothing else. Pick by how many slots are wrong.
The "Swap single piece" alg on SpeedCubeDB only moves the inner wing. For the two-slot pair swap your step 3 drawing shows, the wide-move version below does both slots in one go; it is the same alg with 2-3Lw/2-3Rw in place of 3L/3R.
4Orient one edge
These act on the front edge only. Flips always come as a symmetric pair (both outers, both inners, or all four), because a flipped wing is really the mirror wing swapped into its slot. "Wing parity" and "middle parity" are the same alg on different layers.
+3x3 stage parity
Both edges solved but the last layer refuses: these are the big-cube parities, in the layer notation that actually works on 6x6 (your r / l = 2-3Rw / 2-3Lw).