For my use case (small datasets), query execution feels instant. No perceptible lag between user action and result. Though I haven't done formal benchmarking.
The initial load time is the bottleneck, not query execution. Whether that's the transpiler or just the Go runtime overhead I can't say without proper benchmarking.
I explored the idea of shipping a full backend to the user using WASM-compiled Go, SQLite and OPFS. The whole thing is delivered via a URL but runs entirely on the user's machine.
Total global initialization including Go runtime startup and SQLite connection takes ~280ms on an M1 MacBook Pro.
So well under a second end to end with a warm cache.