Autohinting on large, complex fonts (e.g. several of the Mandeubsida Batang series) sometimes creates overlapping hints that fail fontlint. Some progress already made on this - in particular, catching special cases that were formerly causing NaN values to propagate into the hints - but some instances of the problem remain. They seem to be fragile, appearing or not in a given glyph depending on stuff like the sequence of other glyphs that were hinted, suggesting a variable-initialization problem.
mskala 
Tsukurimashou Font Family and IDSgrep
Autohinting on large, complex fonts (e.g. several of the Mandeubsida Batang series) sometimes creates overlapping hints that fail fontlint. Some progress already made on this - in particular, catching special cases that were formerly causing NaN values to propagate into the hints - but some instances of the problem remain. They seem to be fragile, appearing or not in a given glyph depending on stuff like the sequence of other glyphs that were hinted, suggesting a variable-initialization problem.