Looking at this again, the following issues are present:

• This will add too much runtime overhead, by continually re-computing this for every textLayer element. While that could possibly be solved with caching, there's a couple of other issues as well.
• Regardless of the minimum font-size set in Mozilla Firefox, this function always seem to return 1. Given that Firefox is the primary development target, we unfortunately cannot accept a patch that doesn't work in Firefox while also adding (general) runtime overhead.
• The helper function doesn't really support an unset minimum font-size (which is the default in browsers), since it'll always fallback to return 1 even when that's not correct. At very low zoom levels the textLayer may contain elements with a font-size < 1px, which this patch thus won't handle correctly (and essentially break).

All-in-all, this patch unfortunately seems a bit too far from correct/working; hence let's close this PR and keep the issue open to track the problem.

OK, thank you. I will have a think.