from a

pantsd

perspective, some amount of sub-file incrementalism is possible, because the resulting parsed structures will be identical.