I've got a setup with a single master and two branches. Unfortunately, the network connection to the master is over a 3G wireless network. It's not terribly slow, but the latency is pretty high.
Replication is working well and has been for years, but if there's a network outage, recovery takes much longer than it should. A 1 hour outage can take 30+ minutes to catch up. Outages aren't common, but can last days.
I suspect that the reason may be that while the amount of data is small, there are many transactions (several thousand per hour each of about 200 bytes) and each transaction incurs a substantial amount of network overhead.
Any ideas on how to improve this situation? I saw the pgq_lazy_fetch setting, but I'm not sure this does what I need and looking at the source, it already defaults to 300.
Using postgreSQL 9.5 and skytools 3.2 on Debian Jessie. Master database is actually floating in the middle of a large estuary.