I did this mod to my 1.06 P5A board with a 1k resistor and it works too (with K6-III+ 400 atz). I have found the same issue with the cpu voltage at 2.0v the voltage regulator hasn't signal (0 v) but with 2.1v the system does boot up. I've compared to the rev 1.04 board and the memory performance is looks like little bit different, due to the onchip tag (or the newer northbridge revision), the read and latency is the same, but writing is slower, while copy is faster.
Anyway the other performance (like Quake timedemos, 3DMarks, Superpi) is almost the same, 1.06 board wins with few points but within the measurement error limit.
I've tried a vanilla K6-III 400 (which is in before) and the board and everything seems ok.
I didnt perform a long-term stability test (just few hours Quake1/2/3, 3 (few rounds Openarena under Linux) 3Dmark2k/2k1 loops and prime95), but it was rock stable even on overclocked (112) fsb (I've tried under Dos, Win95OSR2.1, 98SE, 2K SP4, XP SP1, Linux Mint 9 Fluxbox).
Update: I confirm that the mod isnt affecting the memory or other (cpu, graphical etc) performance i've check it with or without the mod and the results are the same. Also the normal cpus works after the mod.
I have another P5A rev 1.06 like this, but it has 1 MB onboard cache and tag chip (I've checked with wpcredit and strangely both tagchips are active), the northbridge revision is the same (G) but little bit older ('99 vs '00) , sadly this board is only stable in Dos and Win95/OSR2.X, newer op's like 98, 2000 or Xp are resetting (like someone mentioned before) if the system is idle (under Linux even during the system startup), under load (and normal cpus) it seems stable. Stability is little better if prevent irq sharing (manaully add irq in bios) and increase the pci latency timer to 64, but not solve it, so the mods are half success 1 off 2 prefectly stable.