It really isn't. Everyone hopping on to the next fad-air-trak-site.io is going to reproduce the same problem. Distributing the data over webtorrent from the beginning will make data access democratic.
I'm genuinely curious but isn't Webtorrent just using WebRTC to join a Torrent Swarm? Torrents are fundamentally immutable, the identifier is a static hash of the content of the torrent. That would mean producing a new torrent for each new data point or chunk of data points only to then submit that hash to a WebRTC based connection to again fetch torrent content?
Genuinely curious, I'm interested in how torrent swarms can be used for novell applications.
If you are going to log, I'd look into CF offerings.