It does not intend to wait for a full file scan before starting file processing. That would slow things down.
Channel Pipeline was my attempt (maybe wrong) to document how data flows through the processing.
There might be a heuristic possible that does a last second check for file changes, but it’d need writing.
A file could grow or shrink while you’re reading it through, so maybe you take the final read size instead.
Problem is I don’t know if you know what size the size-estimator originally had. You’d have to read code.