Slur curvature must follow the same stem-opposite convention as ties, and must be voice-aware in multi-voice writing

In single-voice writing, a slur curves on the side opposite the majority of its notes' stems. In multi-voice writing, each voice's own slur must clear that voice's own stems/beams without colliding with the opposing voice sharing the stave.

CONFIRMED: `Curve` construction (lines 1179-1183) uses a FIXED `y: 20` (positive, i.e. the curve bulges DOWNWARD in SVG's top-down Y axis) for every slur regardless of the underlying voice's stem direction. A positive/downward offset is only correct when the majority of stems in the span point UP (curve-opposite-of-stem convention); a passage whose stems point DOWN (e.g. `rv.stem === -1`, the forced Alto/Bass case in SATB, line 214, or a high-register melody's natural pitch-based stems-down) would need a NEGATIVE offset (curve bulging upward) and instead gets the same downward bulge as an up-stemmed passage — since slurs are 'voice 1 only for now' (line 1173), this risk applies to whatever voice1 actually is in a given lesson, which is roughly half of all real melodic content by stem direction.

Gould (direction convention, mirroring the tie rule); for multi-voice, Gould's SATB chapters and Ross specify each voice's phrase marks must clear both its own stems and the opposing voice on a shared stave.

This rule needs the notated content and the rendered result together — the score says what should happen, the page says whether it did.