In the late 1990s / early 2000s, I did a lot of remote database replication / sync web applications with Sybase database products that were quite mature and excellent. You could actually do syncs on web files (e.g., between local sites running on laptops and sites live on web) by storing web assets in the database.
And, I imagine that the techniques and policies used by those Sybase apps are what people are going to have to recreate for really robust HTML remote storage.
I am sure some frameworks ala jQuery will eventually come into being that let you just set some parameters in a JS config file, and have it handle everything.
But, until then, people will need to figure out partitioning of data, and which partitions sync in which directions (cloud to remote only, two-way, remote to cloud only).
And then there are some partitions that can be state-based-sync’d (just sync to current state–this is like what a VCS does) while others might have to be transaction-based-sync’d (replaying all of the non-sync’d transactions, in order).
And, then of course, you also need a conflict resolution mechanism and error correction for failed syncs!