Speaking of 2-part developers, I thought about how to make PC-512 Borax in two parts. First, I suspect many people will be reluctant to buy and use glycol, and would prefer that both solutions be aqueous. But that raises the problem of the Fenton reaction. It occurred to me that the Fenton reaction will probably not be a problem if part A is (1) highly concentrated, (2) mixed with distilled or DI water, and (3) stored refrigerated, which should greatly slow down the reaction.
The other problem is that borax has a low solubility limit. However, sodium metaborate (tetrahydrate) has a much higher solubility limit, so if you could reformulate this developer with metaborate, having the same image quality, then the following could be a 2-part concentrate to be mixed 1+5+44:
Part A: (1/50th of working solution. Store in refrigerator.)
Ascorbic acid .............. 24 (need to verify that this is low enough to not precipitate when cooled)
Phenidone .................. 1
Sodium metaborate .... ? (make pH of this concentrate 6-7)
DTPA-5Na .................. 1 (if available. Or TEA and salicylic acid?)
Distilled/DI water to 200 ml
Part B: (1/10th of working solution)
Sodium metaborate .... <= 163 (below solubility limit at 0 deg C ensures no precipitation in winter)
EDTA-2Na ................... 5 (not needed if distilled/DI water is used throughout)
Water to 1 L
says the Fenton reaction is most efficient in an acidic environment (pH 2.8-3), so we want part A to be neither acidic nor alkaline. Thus, I suggest that part A be neutral or slightly acidic.
Part A would be kept cooled, and can be poured into the working solution when cold (no need to warm it up) because its amount is tiny.
The EDTA in part B is to prevent precipitation with hard water in both the concentrate and working solution.
I think mixing 1+5+44 would be hard for most people to understand, but the following rephrasing is easy to understand: "You want X ml of dev. Add X/50 ml of part A and X/10 ml of part B. Add water to the X mark."
This idea hinges on switching to metaborate, and a prior poster intends to do just that. I think the metaborate option is worth exploring.