>>18094321
Now, a 2 vowel system which is close to *o ~ *e can be seen in a language like Abkhaz.
https://en.wikipedia.org/wiki/Abkhaz_language#Phonology
Abkhaz has *ɑ & *ə. This is open-back & mid-central. In other words the vowel dichotomy is diagonal instead of horizontal.
Supposing PIE *o was actually /ɔ/, a minor vowel shift ɑ > ɔ and ə > ɛ could produce the PIE system. But why would it shift? One reason would probably be the rise of *[a] which was just */ə/ colored by a laryngeal. *a & *ɑ are so close that */ɔ/ can restore some contrast. PIE *e /ɛ/ is much more frequent than *o /ɔ/, and this would make more sense if *e originally came from an unstressed vowel like */ə/. That would make *o the descendant of the original stressed vowel.
Finally, it's possible this shift /ɑ/ > /ɔ/ was mainly happening in final syllables first for whatever reason. Then preceding syllables were affected by /ɑ/ > /ɔ/ due to vowel harmony with the following syllable. It's actually not immediately obvious /ɑ/ > /ɔ/ was a global PIE vowel shift considering only some IE languages show *o > /o/. Others like Germanic and Sanskrit show *o > /a(ː)/.
If vowels shifted due to vowel harmony, it's entirely possible it was a conditioned change and only some words were affected, even if "some" is an overwhelming majority. (*atta had no vowel shift).
Would vowel harmony be blocked in *táwros? I of course do not know. It may even be difficult to discern the original conditioning factor since it may be expected that many root vowels were analogically replaced by ablaut. But *táwros isn't subject to root ablaut and that may have helped it retain an archaism.
My point is not that I know *táwros isn't a borrowing. Of course it still could be but the issue is the development of the PIE vowel system is poorly understood and has not been described systematically. We need an understanding of the origins of PIE's vowel system to make a competent judgement.