The evidence of 4 base systems, plus several quite scattered and different 'subtractive from 10' systems, suggests that Austronesian number systems may have evolved individually through separate stages in many different areas, ,just as 'Papuan' , and many other languages appear to have done.
They could then have invented or borrowed new words for increasing needs to count exchangeable agricultural or fishing surpluses, and later again adopted very widespread loanwords with more contact and real trade, perhaps long after proto-Austronesians or proto-Oceanics were actually speaking those languages.
(Perhaps, like the Kilivila (Trobriand) chief who got to 9000 and 10000, for counting shells, and ran out of options, they just invented new words on the spot).
Going through New Guinean language records, it's very obvious that the new decimal Tok Pisin has influenced modern speakers very quickly, obliterating earlier recorded systems, at least in the higher numbers. The overlaying process is ongoing, and very visible.
In my Filipino village, everyone now uses Spanish numbers for trade, and nobody can tell me the 'real Surigaonon' for 10, any teens, or 20 up, except 'gatus'=100, which is still used in fishing and agriculture. But the 'native' system was decimal anyway, so there's no radical system change.
----------------------------------------------------------
It should be quite possible, then, to infer multiple overlays of newer systems on old.
Tongan may be an example:
10 = hongofulu
20 = tekau
50 = nai rima avuru (why has hongofulu become avuru?)
(this turned out to be a very definite SNAFU - situation-normal- all-fucked-up.. The Tongan 50 turned out to come from Uruava in the Solomon Islands)
The only An languages that seem to have preserved traces of apparent original number systems are out of the mainstream:
Formosa, Ilongot, Borneo, Sumba, Flores, Timor, SW Maluku, Micronesia, and all of Melanesia south of a fairly definite line.