Yuzu repeated 2T in his Short, got no points on 2T and 3Lz-2T, so it's not just repeating a triple, maybe for repeating any jump in two separate passes? When was this rule first introduced? It seems skaters and commentators alike are not very familiar with this rule.
It's been around since the 1991 season (same year that figures were completely gone, coincidentally).
In the 1970s and 80s, the SP rules required a solo double axel, a specified solo double jump, and a specified double jump in the combination, so the only jump the skaters got to choose was the other (usually triple) jump in the combination.
In 1989 the SP rules loosened to allow men to choose any triple and ladies any double jump out of steps, and the combination jumps were also free choice (one double and one triple, or two triples for men or two doubles for ladies at that point).
So by 1990, several of the top guys were doing 3A+3T combo, solo 3A as the jump out of steps (usually with few steps), and the required double axel. The ISU decided it wasn't a good idea after all to let skaters do axels for all their jumping passes (or solo 3Lz and 3Lz+2T combo, as Paul Wylie and maybe others did instead), so they introduced a rule that it was not permitted to repeat the same jump in more than one jump pass of the short program.
The rule should be, "You can't do two triple toes" instead of "you can't do a second triple toe." That way, if you violate the rule and do two triple toes you lose credit for the lower-scored rather than for the one that comes temporally second.
That would make sense.
It would also make sense for Zayak violations in the free skate (which is what the official "Zayak rule" applies to), to throw out the lower-valued rather than the later element that contributes to the repeat violation.
Or you could just use common sense from the get-go and give credit for the 3Lz, which was, after all, properly done in combination (in actual fact if not in ISU-speak).
That would make less sense, because a jump combination is a required element in the SP, and a 3Lz is not a combination.
However, this year they are throwing out single jumps, or lower-valued double jumps of 2-2 combinations, in short program combos that don't meet the requirements for that short program, and calling the element Jump that counts+COMBO. So in this case they could call a 3Lz+3T as 3Lz+COMBO and give base value for the 3Lz, with -3 GOE.
Any jump coded as +COMBO, whether because the other jump actually executed in the combo didn't count (wrong number of revolutions, or repeat of solo jump as suggested here) or because it didn't exist (if the skater fell on the first jump of the intended combo or otherwise was unable to execute the second jump), gets automatic -3 GOE. That's the only automatic GOE remaining.