Kerning of accidentals in key signature

I couldn't even guess the reason, but LilyPond has an internal inconsistency.

Compare the kerning of the seven naturals when they stand alone (which appears very nice) with how they are squeezed when they appear before the key signature change to seven sharps (distinctively different, and not very nice at all).

Key signatures and chords play by very different rules in MuseScore. The "kerning" in key signatures is done basically the same way for each accidental regardless of font, using a pre-calculated (by eye / tiral and error, I assume) value. "Kerning" of accidentals in chords is done much more precisely. The rule is simple in principle: there is a setting (Style / General / Note / Accidental distance) that controls the minimum distance between accidentals. In this case, it's measured from the lower right corner of the G# to the upper left of the C#.

*However*, there is a slight complication in that the rule for creating "columns" of accidentals by octave (the two C#'s in this example) takes precedence, and we can't kern between columns. So it's really the low C# that is preventing the high G# from being tucked in closer. The column formed by the two C# forms an impregnable barrier against kerning.

In other cases, though, it works just fine. Only accidentals that occur in octave columns are prevented from kerning.