My comments in this reply might be moot for now in as much as the code now says what it says, but I believe the original intent of this section has been substantially bastardized if we were to follow the strict reading Wayne laid out in post #40.
If you go back to the 2011 version of this section (then 310.15(B)(7)) it applied only to 120/240 volt and simply permitted what was shown in the table. Implied were two things: (a) that the section actually modified the NEC ampacity of the conductors, and (b) that all other components in the service would have the rating for the applicable row in the table. I believe that the idea is that a 120/240 feeder really only has two current carrying conductors, and thus an increased ampacity is justified (as compared to the regular table for up to 3 CCCs ). (Someone asked a question about this on this forum in recent years, but I haven't dug through the revision documentation on the change to include 208V to verify. It just makes sense to me.) In any case, the idea that under this section one could protect 4awg cu at 100A, but not use it on a service with a calculated load of greater than the otherwise applicable 310 ampacity, had no support in the way things were written in 2011. Also, the broad permission to use the table implied that adjustment and correction could be ignored. For what it's worth, as far as I can tell this rule had not substantively changed for decades prior.
2014 gave us the revision to "an ampacity not less than 83% of the service rating" and deleted the table. I believe the motivation here was to remove the implication that adjustment and correction could be ignored. I also believe there was no intention to otherwise substantively change the rule. However the way they went about it introduced two new ways of interpreting the section that I believe were unintentional and make no sense from a physics point of view. The first is that the 83% adjustment doesn't apply to the calculated load. The second is that other service conductor components not mentioned in the table, such as meters or MLO panel bussing, also get to take advantage of the 83%. Explaining why they make no sense would make this reply too long, but to me they make no sense.
2017 gave us the inclusion of 3-wire 120/208 feeders. I'd have to look at the substantation, but I believe this was the result of lack of institutional memory, i.e. everyone completely forgetting why 120/240 circuits should get a special allowance here (i.e. one less current carrying conductor).
In 2020 the table was put back in with the caveat that adjustment and correction factors not be applicable. (This is also when it was relocated to 310.12). The phrasing "Table 310.12 shall be permitted to be applied" is sufficiently vague that one could ask whether, in this situation, we are modifying the NEC ampacity of the conductors again. Whatever. It's a mess.
The solution would be to re-write the section so that the ampacities in the table are permitted to be used as the starting point before adjustment and correction factors, and to otherwise restore the section to more or less the way it was in 2011. In other words, make it plain that the NEC ampacity of these conductors is modified in this situation. The table would no longer refer to rating of the service, but just conductor size and ampacity, like any other table. And the phrase 'service rating' and similar would disappear. Which would, blessedly, also make the section shorter again.