For CCPs I have never found a remote method that replaces standing at the step, so I treat remote work as verification of the record trail rather than of the control itself. What does work well from a desk is picking a date at random and following the monitoring record for that shift through to the verification signature and the calibration record for the probe that produced the number, then asking what happened the last time a limit was breached and checking that the deviation record, the product disposition and the corrective action all line up. That tells you fairly quickly whether the system is being run or written up afterwards.
On digital twins and monitoring platforms generally, the model is only as honest as the instrument feeding it, so I still want the calibration trail and one physical check against it. The other question I ask early is whether the critical limit was validated in their own process or lifted from a code of practice, because that one bites the same either way, remote or on site.