For years, what I was mindlessly doing instead: applying a negative margin with “random” values of 0.2em-0.3em for cases like this, which, coincidentally, almost the same!

Intuitively, and by testing, I was pleased with these values all this time, but having a proper method on one hand, and a confirmation that this type of adjustment results in what the fallback ends up being in the end is nice, haha.