That's the problem with "partial seeding". only getting not-ALL of the transfer files.
More overhead because you have to tell EACH client what you DO HAVE. The default interaction is "have all" or "have none" when joining a swarm. As you get more of a transfer that bitfield fills up, but theoretically you're only sending incremental updates to each peer you connect to.
This is also why Tixati uses more RAM than other clients.
It essentially caches all the status for EACH peer in EACH transfer you have to save time later if they reconnect for instance. Or if your i-net drops out and can reconnect to them later. It can resume from "last state".
https://support.tixati.com/transfer%20properties%20-%20options