Keeping continuous replication without shipping another binary is the appealing part here. WAL streaming in pure JS should also make this easier to run in restricted environments where spawning a subprocess is not an option.
The distribution argument is the convincing one. A sidecar binary is enough friction for most small teams to just pick Postgres, but an npm install inside the app process removes that excuse. Keeping the LTX files byte-compatible with the Go version is a smart choice too.
netproxy
The honest limits section is what makes this post trustworthy. The single-writer caveat with a shared prefix is exactly the kind of footgun that would bite in production. The streaming numbers are reassuring too — memory bounded by the page buffer rather than data size is the whole point. Curious to see the LZ4 work land.