For example, the entire station may be disconnected from the grid, but somewhere in the system is a Siemens device that for whatever ungodly reason needs to access the internet, which works when the grid as a whole is operational but fails when it can't get cell service.
And even if those stations work correctly, you still haven't verified that the actual plan is operational because you haven't actually done the black start; perhaps the grid divisions are incorrect and the first load you try to apply will brown out the start source.
It's the same reason that datacenters don't just run the generators every month, but actually switch over to generator power and disconnect from the grid to verify it works (and if it doesn't, hopefully the batteries last long enough to get back to the grid!).