Author here. The single-node restriction for Litestream was one of the main reasons we started LiteFS. There isn't a way to handle streaming backup from multipl

Author here. The single-node restriction for Litestream was one of the main reasons we started LiteFS. There isn't a way to handle streaming backup from multiple nodes with Litestream & S3 as SQLite is a single-writer system and there aren't any coordination primitives available with S3. I agree that many of the SQLite cloud offerings introduce the same network overhead. With LiteFS, the goal is to have the data on the application node so you can avoid the network latency for most requests. Writes still need to go to the primary so that's unavoidable but read requests can be served directly from the replica. The LiteFS HTTP proxy was introduced as an easy way to have LiteFS manage consistency transparently so you can get read-your-writes consistency on replicas and strict serializability on the primary. That level of consistency works for a lot of applications but if you need stronger guarantees then there's usually trade-offs to be made.