The closest formula would be the one in the patent.
However the nagging issue when scratch mixing anything with ascorbate is preventing the dreaded Fenton reaction (premature oxidation of ascorbate). You need the right iron sequestering agent, or else it will be difficult to predict shelf life.
Alternatively, D-76/ID-11 is very close to XTOL in working characteristics. XTOL is my favourite developer, and rather importantly in the history of black and white chemistry, was basically the first (and last) developer to improve on the D-76 speed-grain-sharpness balance. But the improvements are pretty small, so all things considered it is far easier to scratch mix D-76 than XTOL and get nearly/virtually the same results (same tonality, very slightly different image structure).
A bit late reply, but here it is...
I've tried a modified patent formula, with very good results. I didn't have DTPA-5Na, so I used sodium hexametaphosphate instead. I also didn't have any kodalk, so I used borax, with the addition of the right amount of sodium hydroxide. Sodium hydroxide was also used to neutralise the ascorbic acid I had, instead of sodium ascorbate. Finally, I don't have any Dimezone-S, so I just used the same amount of Phenidone. That's not a scientifically calculated value, I just assumed that it will work fairly well and had no clue whatsoever what an equivalent amount would be in terms of activity level. For the most part, I used deionised water and always kept full strength developer in plastic bottles, filled to the rim. The pH of this developer is a bit on the low side, because DTPA-5Na increases alkalinity, but a bit of sodium hydroxide will correct it. I never had any failures with all the batches that I made. 6 months old developer that stood unused in a soda bottle worked perfectly.
I then sourced some DTPA-5Na and mixed another small, 600ml batch. This time I used tap water and keep in mind that there is at least one iron pipe in the building. This time, the only practical difference between my mix and the patent formula is the substitution of Dimezone-S with Phenidone and pH was found to be 8,21, practically dead on. This time, I decided to reuse this batch and increase development time according to the datasheet instructions and it worked perfectly well. Actually, the last time I used it was 9 months after mixing this batch and 6-7 films had been developed with it, very close to the limit reported in the datasheet. The bottle looks ugly with deposits on the walls, but the solution is nice and clear, perhaps with a slight hint of yellowing, just like my replenished Xtol did many years ago.
So, am I lucky for not having Xtol (commercial or homebrew) die on me? Maybe, but I always had the suspicion that Xtol failures are far less frequent than what we think and basically overreported. In any case, a clip test before committing any film to it is a good idea and doesn't take too much time. There are no shops selling photographic chemicals around here, I need to order everything I need. I'm not a high volume user either, so 5l of Xtol is a tad too much. I'll keep mixing my homebrew Xtol, the results are very nice and it costs very little.