[AutotelicBrown]

The main problems I see with the current system is that SAP gains after you achieve 40 (or any arbitrary top tier) are essentially useless as SAP deductions are unlikely to happen, and that prolific steppers aren't able to push a few more files into the game if the work is truly top-notch.

What about introducing a few more tiers with an added change that, if you want to make use of them (i.e. submitting more than 5 files), you have to go through a progressively harsher SAP bonus/deduction table up to a soft limit where a 10/10 file would net 0 points and anything lower gives a proportional deduction to SAP.

Just as an example, a table of SAP changes according to the number of files being submitted for the batch and obtained average score (PS: I know those values are bad):

--------| 1-5 files (current values) | 6 files | 7 files | 8 files
9.5-10. | +5 | +4 | +2 | +1
9.0-9.5 | +4 | +3 | +1 | 0
8.3-9.0 | +3 | +2 | 0 | -1
7.5-8.3 | +2 | +1 | -1 | -4
6.7-7.5 | +1 | 0 | -3 | -7
1.5-6.7 | 0 | -1 | -5 | -10
1.0-1.5 | -1 | -2 | -7 | -13
0.0-1.0 | -2 | -3 | -9 | -16

The benefits from that would be:
-Doesn't change what already works
-SAP gains are always useful as you can 'spend' them sending more files
-Actual incentive to make top-notch files
-Don't hinder experimentation as in the worst case scenario you are at the base 5 files limit

It softens the saturation problem as the stepper is more likely to spend SAP submitting a higher number of files. In that case the stepper wouldn't be able to sustain a high number of submissions per batch.
Unless the stepper can consistently push godlike scores on a reasonable amount of files under a volatile judgement system, in which case I don't see why the game shouldn't get a few more top-quality files than usual.

The idea is really raw, I have no idea what would be good values and I don't have a holistic view of the system, but I guess it doesn't hurt to share it.