Thankfully, I’ve yet to have to do a full restore “in anger” of synapse/psql, so I’d have to answer “It worked in testing” 
To answer the implicit question though “would restoring the db files even work, instead of a DB dump?”; IMO this is tested every time the docker swarm load balances, or updates, as these processes involve a complete replacement of the psql and synapse containers.
To close the thought, I am in complete agreement that you should absolutely make sure your backups actually do work when restored, and have some sort of Data Recovery (DR) plan prepared in advance of when things go bang. How thorough that needs to be (and if/how often you run through DR tests) depends on how important your data and/or your application uptime is to you 