There are lots of ways to implement this more efficiently, but for the first pass, re-running any currently "live" query on a tx-report (against a user-security-filtered db), diffing the new result with the previous one, and sending the new datoms to the client should suffice.
Once we see what that looks like as a practical API and what requirements it has at runtime, we can start to prioritize a simpler set of "live" primitives.
There are lots of ways to implement this more efficiently, but for the first pass, re-running any currently "live" query on a tx-report (against a user-security-filtered db), diffing the new result with the previous one, and sending the new datoms to the client should suffice.
Once we see what that looks like as a practical API and what requirements it has at runtime, we can start to prioritize a simpler set of "live" primitives.