Slices remain — the caller must keep the on-demand frame loop alive (the per-frame pump is the only thing driving the stream forward).
The active slot changed on this call: the newly-written buffer is now the one to draw. The caller must push activeSortedIndexSlot to the node's materials.
This call made NO progress: the previous slice's upload has not been
acknowledged (its onUploadCallback has not fired — the mesh was
hidden, frustum-culled, or otherwise not drawn), so advancing would
union the next slice onto the still-pending one and grow the pending
upload without bound (issue #715). more stays
true and flipped stays false; the caller must NOT spin the
on-demand loop on a stall (a subsequent DRAWN frame consumes the
pending slice and the next pump advances). Always false on the
WebGPU single-slice path, which never stalls.
Outcome of one pumpSortedIndexOrderingApply call.