If everyone wants to fill up their tank at once, the gas stations run out of gas. This doesn't happen with any regularity, which I take as an indicator that the scenario you're describing is very rare.

The gas system has buffering and slack at many levels. Cars store more mileage. Gas stations have a buffer. Gas stations can request more trucks come refill much easier than a grid can be upgraded. I don't think you proved anything.

And yet runs on gas stations happen during crisis times and they do in fact run out of gas.

Precisely, even with that significant slack, it still happens. How frequent will be the overloads in a system with no energy storage capacity, which is sized to handle the existing load peaks and already has infrequent overloads?

The average commute in the US is 50 Km per day and if that is the average it is reasonable to assume some communities will be highly correlated above average, for example if they are 40-50 Km away from a large economic hub, a large proportion of residents will drive the 100Km roundtrip per working day. The lower price of electric comute will prompt many residences to have two electric cars driving the average every working day, for a total, again, of 100Km = 20KWh. So double digit is not exceptional, it will be the average, an extra 3KW of power for each household for the entire night, assuming an outstandingly smart grid that can perfectly level off demand.

Combine that with rare events like hot nights and large movements, psychological reactions like panic, and you have an unreliable local supply that might trip every few months without expensive upgrades.