splitAtAntimeridian works from which COPY of the world each longitude is
in, so it needs a series that runs past ±180 rather than jumping back.
greatCirclePoints produces one; a satellite ground track does not — SGP4
reports every longitude inside (−180, 180]. A track leaving Kamchatka at
+179.9 and arriving in Alaska at −179.9 therefore looked like a single
359.8° step inside one world, no split was made, and the polyline drew
the chord: a straight dashed line clean across the map on every crossing.
The series is now unwrapped first. A jump of more than 180° between
consecutive points is that wrap and nothing else — no real path steps
half the globe between samples — and the unwrap is idempotent, so the
great-circle caller is unaffected.
Checked on three series: a normalised crossing now yields two segments
meeting exactly at ±180 where it used to yield one chord; an already
continuous 170→183 still splits at 180; a track that never crosses comes
back untouched.
rouggy
d0d29659cb