2/
Normally there's only one controlling node in charge of many others, but various "interrupt"-type situations call for any of the others to spontaneously emit a packet. These spontaneous transmissions cause bus contention if multiple nodes try to transmit at once. Since my protocol has checksumming and retransmission, I don't need to handle actual multiple access, I just have to detect this contention and have at least one side abort the transmission attempt and try again later.