The voice now converts the file's rate to the host's at note-on instead of assuming they match. Stage times resolve at the host rate, the filter solves at it, and the bake follows. Same-rate renders are unchanged bit for bit.
