Many Javanese words have a schwa in their first syllable: təlu (three), pəkan (market), dəmak (gift). In Old Jav texts these usually written as <təlu> or <tlu>, with the unwritten first vowel pronounced as normal. But in the oldest inscriptions I'm finding a weird third option...

Instead of schwa, the vowel is written <a> and the next consonant is doubled: <tallu>, <pakkan>, <dammak>. This leads to some forms that, in my very subjective opinion, look quite un-Javanese: <inammassan> for inəmasan (gilded)

Is this just a spelling artifice, like doubling consonants after the letter <r>: e.g. /varṇa/ written <varṇna>? Or does it represent a phonemic reality? In final syllables, /a/ often weakened to /ə/, e.g. pasak > pasək, dalam > daləm. But I don't think that's what this is

If spellings like <tallu> and <taŋŋah> represented genuine early forms of the words /təlu/ and /təŋah/, wouldn't we expect the double consonant to remain? Moreover, the reconstructed Proto-Austronesian forms of these words have schwa with single consonant (*təlu, *təŋaq)

The <tallu> forms are not that common and they seem to be interchangeable with the more frequent <tlu>. Maybe the <tallu> option represents some other linguistic feature, perhaps accent? Any ideas welcome, I'm an amateur at this historical linguistics stuff

The reason I suspect it could be a prosodic thing is because doubling happens in another context. In these 9th c texts, when a consonant-final word gets a vowel-initial suffix, the consonant doubles: ka+hurip+an = <kahurippan>, umadag+a = <umadagga>, kumon+akan = <kumonnakan>

The effect of this is keep the final syllable of the stem closed. In <kahurip>, "rip" is closed, but in <kahuripan>, it is open (ka-hu-ri-pan). Doubling the consonant keeps it closed (ka-hu-rip-pan). Why would you want to do this?