I agree that must-pull is likely overreach. But even in the speedrunning context the other responses are rightfully ignoring for the most part, as a color buffer she seems like one of the safest bets in terms of a unit appreciating as we get more content and units (that is, unless you expect one of those units will be a more accessible yellow buffer).

For some apples-to-apples we can look at:
- yellow ShiroKuro torment: 2023SEP 43% → 2024JUL 96%. Gear and level creep enable Mika to be a sufficient tank for Kuro without dedicating slots to repositioning/obstacles/etc which makes damage luxuries like S.Ui easier to slot in. Note that for the tracking stats on Bluearchive-ranking that I assume you're referring to (and so I'm using as well), Torment has outgrown the 10k cutoff so the stats are more focused on clear speed since kitchen sink comps that can compensate for no S.Ui are no longer represented.
  Also, if you narrow down to top 100 she goes from 98% → 100% usage rate (from arona.ai), so on some level her under-representation comes down to skill/investment issue rather than lack of optimality.
- yellow Binah torment: 2023OCT 35% → 2024JUL 43%. Lack of sufficiently strong gear/levels/defence down units meant that all the torment clears of the prior raid were to some degree kitchen sink comps throwing many teams at Binah to eek out a win. In this case S.Ui was simply not optimal, going from 2% → 85%; she simply was not numerically worthwhile compared to maximizing defense downs on the main damage carry teams and thus had no place at all. In this case the kitchen sink-ery buoyed her overall Torment presence since she was at least some kind of buffer. And then basically reverse for the latter raid: gear & level creep allowing for Mika to (barely) consolidate the tank role on top of DPS makes a luxury buffer like S.Ui optimal and thus increasingly more mandatory as the torment cohort grows.
- yellow Hod torment: 2023NOV 4% → 2024MAY 17%. Unimpressive. In both cases it largely comes down to her not solving the gimmick or being the easiest to use buffer in whatever slots remain. 9% → 65% in top 100 paints a more positive picture in terms of optimality progression, mainly because UE40 C.Kotama consolidates/streamlines the CC gimmick which opens up a spot for S.Ui.
- yellow Chesed torment: 2023NOV 74% → 2024MAR 3%. Originally closer to a kitchen sink approach, where I believe S.Ui mostly doubles as enabling Cherino to be sufficient waveclear and buffer for Mika in the mid-to-low end and numerically just the best buffer in slot for a dedicated single target finishing team on the high end. With extra levels I believe S.Hoshino suffices for Cherino in the latter raid which is why we see a massive drop-off. In the top 100 slice she still sits at 97%→100% for the people who can bother with finagling a single-target dedicated finishing team (pivotally Mika could now 1-cycle the core with some clever deaths to loop her EX a bunch).
- yellow Goz torment: 2023DEC 86% → 2024AUG 76%. Units like C.Hare & D.Aru offer more options for potential second phase teams that clear sufficiently well so S.Ui isn't as mandatory overall. At the extreme malding top 100 end she maintains 100% representation but in seperate rolls, as in the latter raid she's run in the first half to consolidate Reisa as both waveclear and tank.

S.Ui is weird in that I think she has a U-shaped desireability curve. At the middle end she doesn't really gatekeep Torment much, and possibly less so as time goes on. But as plat becomes more competitive within torment, or you're aiming to push ladder hard, she's becoming progressively more requisite. Others already discussed her on the lower end.