(U) F2 U//EO (3/3)
F' L' F B' L D2 L//DR, 2c3 (7/10)
(L2 D2 B D2 B D2 B)//HTR (7/17)
(L2 B2 U2)//M slice (3/20)
...F B' L D2 [L U2] B2 L2 B'...
[]->R B2 R' L (22)
Final: F2 U F' L' F B' L D2 R B2 R' L B2 L2 B' D2 B' D2 B' D2 L2 U'
6/6/8 scramble
53 EOs checked
I looked at these other DRs:
(L2 U' F' D)
(L' F/F' L2 R B' R)//two 2c3s, one of them is 21 optimal (10)
(B) U' D' B
(D2 R U L2 U L)//2c3 (10)
I focused more time on the DR I used because it was HTR-2e2c which was better than the subsets of the other 3 10 to 2c3's I got. Double slice redux ftw

(F U' R2 F)//EO (4/4)
L U2 B2 R D'//DR, 2c3 (5/9)
[R' D2 R D2 R'//HTR (5/14)
D2 U2 R2] U2//M+S slice (4/18)
[]->L' F2 L' R2 U2 R' U2 D2 L2 (19)
...D' L' F2 L' R2 U2 R' U2 D2 {L2 U2 F'} R2 U F'
{}->F' B D2 L2 B' (21)
Final: L U2 B2 R D' L' F2 L' R2 U2 R' U2 D2 F' B D2 L2 B' R2 U F'
4/4/6 scramble
47 EOs checked
My main 'mistake' on this scramble was spending too much time on 5 move EOs. I guess I was mentally biased after the events of attempt 1? I basically had no time left to check all the 4 to RZP from the 2 move EO, and also didn't quite spend enough time on the 9 to 2q that Fush used. I think I found his HTR but with the wrong slice configuration that made the floppy reduction skip not so obvious to see. I also spent a lot of time on the random 14 to HTRs but to no avail.
I had these other DRs:
D2 L' 
U L2 F D2 F2 B' U//0c4 (9)
D2 L'
(F) U2 F2 U' R2 D B//4b2, this is the one that Fush used for his 20 single (9)
(F U' R2 F)
(L F2 D2 L U')//2c3 (9)
L' U' L' D' F'
L R B2 R F2 U'//4b2 (11)
L D2 R//HTR in 14
L' U' L' D' F'
(L' F2 L R' F2 R' D)//4aq (12)
(B2 L)//HTR in 14
(R F' U' B2 D)
R2 F' U2 L//4b5 (9)
(F' D' R2 F' U)
(B' L' B L B L)//4b3 (11)
(R F U' B2 D)
(R) F2 L U2 F2 R//4b2 (11)
The optimal for this scramble is findable:
D2 L'
(F L2 F2 L2 U' R2 D' F')//10 to DR that no one tried because the rzp case just looks bad
(B2 D' B2 L2 U' F2 L2 U2)//finish