Firstly, there are naturally occurring mixed-chirality (alternating) peptides. They are usually circular iirc.
Secondly, no you can't really have larger proteins with both left and right (ignoring glycine). They would not fold into nice helix/sheet strucures and likely just be random coil.
For cells to have mixed populations of all-L and all-R proteins would mean doubling up all the machinery for creating them.
One theory that I thought was reasonable for why there's a monochiral world is that once the arbitrary choice is made (L or R) then that gets 'locked in' by all the machinery around that choice. As in, L 'won'.
This seems obviously true to me. Mixed doesn't work, so as molecules and systems of molecules started replicating one chirality won out. It's just chance and there's nothing magical about the chirality "chosen" by the process.
E.g. a ring structure whose substituents are affected by steric hindrance in the left-handed scheme.
And the path of least resistance was just to adapt and build around it. Once that precedence was set everything became as such. I expect in the earliest stages of life this would have been an immense factor as metabolism was not nearly as sophisticated as we know it today.
And this selective process may have ocurred well before anything we have observed/modeled, and may well be erased. Which is to say I agree, but with the caveat that it was a substrate-dependent mechanism which selected the downstream components rather than random chance.
Totally agree that it is hard to test these things experimentally, or through historical analysis of structural remnants. I understand there have been efforts to model at a system level these ancient metabolic networks but ... then how do you experimentally validate these models?
> might be tricky to distinguish chemically during the protein synthesis process
No, amino acids are bind to tRNA by special proteins that have handiness and can easily distinguish the L and R version. Most proteins can only operate on one handiness of the target molecule.
> making e.g. different codons for left and right-handed amino acids infeasible to implement
No, there are 64 codons and we are using them to map only 20 amino acids and a stop signal. So there is a lot of duplication. Some bacterias have one or two more amino acids or a small tweak in one or two of the conversion table, so it's possible to add more stuff there if necessary.
My guess is that mixing L and R amino acid would break ribosomes. The ribosomes read the mRNA and pick the correct tRNA and connect the amino acid that the tRNA has. I guess that the part that makes the connection assumes the correct handiness of the amino acids.
Going down the rabbit hole I found https://en.wikipedia.org/wiki/Nonribosomal_peptide that explains that some peptides (that are like small proteins) are formed by special enzymes instead of ribosomes, and some of them have D-amino acids or other weirs stuff.
Ah, neat. That was the step where I worried about coding being infeasible, too, coding for R amino acids wouldn't do any good if you couldn't distinguish them. I did know there was plenty of room in the encoding scheme.