The words e-ku-ut-ta and e-ku-ud-du do not sow voice assimilation.
Although if some secondary feature was assumed, a phonemic contrast would be possible in spite of the assimilation, but claiming that would be quite ad hoc.
causative ḫi-verbs in -aḫḫ- usually take the first person singular desinence -mi instead of the appropiate -ḫi, to avoid the quadruple ḫ. This makes sense if ḫḫ is interpreted as [ħː], since a quadruple ḫ would be impossible to articulate, and it would be simplified to just a fortis ḫ, which would be identical to the third person. In order to disambiguate these forms, the first person singular desinence -mi is borrowed from the first conjugation.
Although ḫi-verbs in -aḫḫ- usually take the first person singular desinence -mi instead of the appropiate -ḫi, there are at least two attestation of a word with its proper desinence showing a quadruple ḫ spelled with a dummy vowel as aḫḫaḫḫi, this can hardly represent two fortis consonants, but rather two true geminate pharyngeals.
The Hittite word zuppa- in Akkadian as zu-pá-nu "a metal container". The stop always attested voiceless, and notably, never geminate.
The Hittite word 𒄩𒇻𒅗𒀸(ḫa-lu-ka-aš/ħluːgas/) is attested in Akkadian loan words as ḫu-lu-ga-an-nu-um and ḫi-lu-ga-an-nu-um. Since Akkadian did distiguish voice and voiceless cuneiform sings, this reveals that the velar stop was voiced. These could not follow any Hittite orthographic convention, since these loans predate the Hittite period.
However, it may be that lenis "k" had voiced and voiceless allophones. It is common for voiceless stops to become voiced in intervocalic position. Or, it may be that just as fortis consonants were lenited by Eichner's laws, lenis consonants could have been voiced by the same process.
The city URUTarḫuntašša is written in Ugaritic as trġnds, were "ḫ" is voiced. It is surprising that the "t" is written voiced, this may be because of a possible lenition of all stops from Luwian.
However, neither of the obstruents is expected to be voiced, so there is no reason to think this Ugaritic loan is representative of the true pronunciation.
The Ugaritic loan ḥtṯ < ḫattuš "silver" does not show any gemination, and shows that geminates were voiceless indeed.
The Ugaritic loan dġṯ < duḫḫuiš "silver" does not show any gemination. It showes that the initial stop was voiced, which is in agreement with it's etymology "dhewh₂-".
It is however unexpected for "ḫḫ" to be written as "ġ" if it's supposedly voiceless in this framework.
