Less Wrong team: what seems to be going on is that you’re using TypeKit’s subsetting feature to specify a quite restrictive subset for the Warnock Pro font, which is excluding characters such as ‘ë’. Thus the copy of the font file that’s being served to visitors of Less Wrong doesn’t include that character.
As a result, ‘ë’ (and other characters with diacritical marks) is being rendered in one of the fallback fonts. Naturally, in the case of some of those fonts, the x-height and other metrics of the fallback font will differ from those of Warnock Pro, resulting in the visual glitch linked above.
This is easily fixed by specifying a less aggressive subsetting config in TypeKit for Warnock Pro.
Ok, I think see the issue.
Less Wrong team: what seems to be going on is that you’re using TypeKit’s subsetting feature to specify a quite restrictive subset for the Warnock Pro font, which is excluding characters such as ‘ë’. Thus the copy of the font file that’s being served to visitors of Less Wrong doesn’t include that character.
As a result, ‘ë’ (and other characters with diacritical marks) is being rendered in one of the fallback fonts. Naturally, in the case of some of those fonts, the x-height and other metrics of the fallback font will differ from those of Warnock Pro, resulting in the visual glitch linked above.
This is easily fixed by specifying a less aggressive subsetting config in TypeKit for Warnock Pro.