On three elements:
(1) client's request, (2) final printing-size & (3) original pic.


2) If the file has to be printed / visualized in low dimension, is unuseful to exceed in details. They wouldn't be noticed or they could even (if very small) seem errors or imprecisions.

3) If the original shot has few details (dark,smallor low-resolution pic) it gets very harder, for us, dto rawn without a referring point.