Chord.Voice(octave, inversion) puts the root back in the bass from the 4th inversion of extended chords: C9 inversion 4 gives C5 D5 E5 G5 A#5 #356

What's wrong

Chord.Voice(int octave, int inversion) (Chord.cs#L529-L563) builds inversion k by adding 12 to the first k chord tones and re-sorting. The doc says inversion k voices the chord with the k-th chord tone lowest. That only holds while every tone above the raised ones is within an octave of the root.

ChordTones() returns tensions as 13–21 semitones. Once the root is raised to 12, it sits below the 9th (14), 11th (17) and 13th (21). So when the inversion reaches a tension, the raised root drops back to the bottom.

Reproduction (against ce31509)

Inversions 5 and 6 of C13 also put C in the bass. The inversion is silently ignored, and the result is root position an octave up.

Trace for C9: the tones [0,4,7,10,14] become [12,16,19,22,14] after four raises, which sorts to [12,14,…].

There is a related smaller problem. The root is built with Pitch.Parse(Root.Name + octave), so a root above MIDI 127 throws FormatException("Invalid note name 'B9'") (Chord.Parse("B").Voice(9)). An upper tone out of range throws ArgumentOutOfRangeException for midi (Chord.Parse("G").Voice(9)). The same kind of bad input gives two different exceptions.

Suggested fix

Take the target bass as tones[rotation], then raise every tone below it by whole octaves until it sits above that bass, instead of raising only the first rotation tones by one octave. Compute the root from MIDI arithmetic rather than by parsing a string. Throw a single ArgumentOutOfRangeException(nameof(octave)) when the voicing leaves 0..127.

Acceptance criteria

• For every chord in the round-trip corpus and every inversion in 0..ChordTones().Count-1, the lowest pitch of Voice(4, inversion) (ignoring a slash bass) has pitch class Root + ChordTones()[inversion].
• Voice with an octave that pushes any tone outside MIDI 0..127 throws ArgumentOutOfRangeException.

The existing InversionTransposeTests cover only triads.