I made two discoveries this weekend.
1. The toe in Delta 400 can be shortened to match Xtol by reducing ascorbic acid (AA) to 3.7 g or lower. D316 uses 4.5 g, and I tested 3.0, 3.5, 3.7 and 4.0 g this weekend. At 3.7 g, sodium metaborate must be reduced to about 1.7 g. This raises pH some to compensate for loss of activity due to reduced AA. Grain is still at least as good as Xtol. BTW, my test-strips contain two frames each, so I can measure two toes in each test, giving me more confidence that I didn't see an anomaly.
2. Even with reduced buffering due to reduced alkali and acid above, highlights and shoulder were identical. I finally realized why: The shoulder is caused by the limited diffusion-rate of the gelatin. I designed my developer to have a higher devtime-to-phenidone ratio than Xtol. As a result, my dev needs a lower diffusion-rate than Xtol (which I never thought about). So it hits the diffusion-limit at a higher density. Hence the higher shoulder. (PE, please correct me if I'm off in the weeds here).
This gives me an idea of how to make a developer that's good for pushing: Reduce that ratio so much that even midtones sag. That will give the toe-area time to develop more while taming overall contrast.
3. I'm happy because I killed the gopher that was tearing up my back yard. Oh, this is off-topic...