You don't. Yet.
Tesla knows the status of all of the stalls in (near) real time - which ones are damaged, which ones have a paired car that's drawing a lot, which cars are almost done and drawing next to nothing.
I'm hoping (and expecting) that one of these days I'll wake up to find that little yellow alarm clock on the top bar, and the next time I go to a Supercharger it'll recommend a stall, and show me which one it is on a pop-up map of the site.
When several are empty and not degraded, Tesla can use some sort of load leveling algorithm to balance the wear. When everything is shared, it'll point to the one with the highest power available.
Tesla is going to have to be digitizing the sites and setting up a method of picking for the FSDC demonstration at the end of the year anyway - why not include it in a general firmware update as a way to streamline and improve efficiency and decrease the hassle of switching from stall to stall?
This can be tied into an automatic sequencing/reservation system for overloaded sites, too - based on order of arrival, it gives you a "place in line" on the center screen - maybe even an estimated wait time if enough of the cars are running on Nav routings.
Another bonus for Tesla is they can use the same data they use for this to automatically detect and report stalls that have issues.
One challenge will be how this sort of system handles ICE'd stalls. Maybe they give you a button that says "not accessible" or something like that in the pop up, and it gives you another stall recommendation?
If a bunch of folks hit the blocked button on the same stall over a period of time, Tesla can talk to someone about towing...