Thread by @PhDniX

So... Japanese native numerals are *weird*. People like to talk about the doubling ablaut system
hito-tu '1' ~ huta-tu '2'
miC-tu '3' ~ muC-tu '6'
yoC-tu '4' ~ yaC-tu '8'

But that really doesn't cover all the odd exceptional behaviour found in counters...

Why:
hutu-ka "2 days" not huta-ka?
why miC-ka "3 days" but not muC-ka but muika "six days"
Why yoC-ka "4 days" but not yaC-ka but youka"8 days"

(and why nano-ka "seven days" while it is nanatsu?)

I see no obvious reconstruction that accounts for those irregularities.

Is there any article that tries to account historically for these unusual irregularities?

muika and youka look like they come from onbin changes from < *muki-ka and *yaku-ka or *yapu-ka/*yapi-ka
But I don't think you can get mikka or yokka from something that involves those stem consonants...

I suppose you could get both *yokka and *youka from < *yopi/u-ka and *yapi/u-ku, if the former continues the eastern dialectal realization yopka> yokka and the latter the western dialectal realization yapuka > yauka > yōka... They would become homophonous in western dialects...