In summary we discussed how Julio's new syncing code is more ambitious than originally planned.
We currently have a few mechanisms for transferring data between nodes:
Periodic Subscription Sync - regularly update content the user is subscribed to
Push - send content to site
Discovery - request content from peers
The push project was making change to subscription sync. There is a new more sophisticated algorithm which requires the server to do a bunch of work to prepare all possible related content, then do RBSR syncing
But we decided that was too heavy and risky for now, in particular if it changes the periodic syncing process.
The new approach is to mimic the current behavior of the TS sync implementation, where the client determines the relevant content and then ensures the server has it. This should include IPFS media files.
Once this completes, we will be able to get rid of some TS code, including the client-side syncing logic and the hm/api/discover endpoint, which is how the TS sync was working.