This is an interesting question because it raises a number of different issues concerning the evolution of the universal code and the protein synthesis machinery.
The main thing that I would highlight is that your question almost assumes that the universal code evolved fully formed and that all codons would automatically be present. Is it not more likely that "stop" would be the default situation and that the universal code (and the corresponding tRNAs) evolved gradually.
The second thing to consider is that the ”mechanism’ for preventing “stop tRNAs” popping up does not need to be a cellular one. What is wrong with good old evolution? Stop codons are not a complete necessity, as the ribosome cannot go on forever and would drop off eventually, but losing them all would be a disadvantage. Conversley, losing a seldom used tRNA gene and gaining a useful “stop” could be an advantage.
I suspect that this all a gross simplification but it would be interesting to see whether anyone knows of any research that tackles these ideas. I did find a paper ages ago that discussed the gradual evolution of the universal code by looking at the sequences of tRNA transferases, but can’t unearth it at the moment.
™ Version 4.2