There were some interesting suggestions here. I strongly prefer repeatable procedure avoiding infrequent agitation. Even if the solution is simple, if it requires a magic hand to make it, it doesn't seem to be a good solution to me.(posted 8170 days ago)If some pyro or pyrocatechin developer can make usable shoulder, I suppose it is easy to reduce or remove the stain after that point to avoid exceedingly lowered highlight contrast. One problem is grain, as John Hicks suggested. If grain can be tamed, it might be a good soup for TMY :-) If grain is unavoidable, I might just go back to Plus-X instead, depending on how much modification the new version receives.
I might have said a little about this in the original posting, but the most likely reason staining developers can help achieving compensation is through local hardening of gelatin. The oxidation products of those agents increase cross linking of gelatin, making it shrunk, harder, as well as stained. However, literature data suggest local hardening requires less oxidation products than staining. A caveat here is that those data were obtained with old emusions and relatively high pH, low salt environment, letting the soft gelatin swollen as much as it liked until development progressed.
"Is it possible to get the higher shadow contrast through reciprocity departure [...]?"
This is an interesting question. Something I never wanted before! Non tabular grain films of high speed gives you great depearture from reciprocity law, but you want to chemically create it. I suggest trying various restrainers. But alkali metal salts of halides are probably poor choice because they will turn into silver halide solvent at high concentration, possibly before you get desired restraining effect. (This may sound strange violation of solubility product principle, but the solvent effect is because of formation of soluble silver halide complex ions, and is subject to fog and solution physical development.) I suggest those restraining dyes should be tried. This, obviously, I have never tried and I don't know how workable the idea is.