Dropbox published an overview of how part of their sync works (Inside LAN Sync, HN comments), specifically how it can take advantage of LAN segments and avoid going outside the network for large chunks.
None of it is terribly surprising but can be nice to see validation and see how others are solving similar problems.
They get an easier time using a central point as a synchronized one-source-of-truth about file metadata and versions, instead of syncthing’s consensus algorithm.