a.) There are no initial or incremental/differential backups - only deduplicated backups.
b.) After stop/start, backup needs to read the whole disks, but backup are still fully deduplicated.
c.) backup data is deduplicated at the server side.
PBS currently use a single http2 connection per backup. A future plan is to use http3, but I do not
think that can saturate a LACP connection. So, no - there are no plans to optimize for LACP.
The chance of a collision is lower than winning 8 lottery games in a row, see
https://pbs.proxmox.com/docs/technical-overview.html#notes-on-hash-collisions