Allow small increase in output chunk size for transform preprocessing (E8E9). Ensure chunk allocations always include oerhead size.