This is generally a good idea when interacting with human drivers and in the outside world where the probability of encountering another self-driving car in such a situation is extremely low. It only breaks in this specific edge-case where there's seemingly nothing but self-driving cars in the entire lot.
It's actually quite a difficult problem to solve. The honk reflex is there for safety - if something that might be a car (based on limited/occluded camera & sensor data) you want to honk just in case, as early as possible. Waiting for the presumed car to get into better view so you can identify it is dangerous as it will delay the honk reaction to a time where it might be too late.
The only way I see to solve that is that all the cars need to communicate and be aware of each other in 3D space, and then be able to predict their upcoming position in the visual/radar field very precisely, so that as soon as it comes into view you need to be able to confidently say that this is indeed another Waymo car and thus inhibit the (now unnecessary) honk reaction. This needs to be very precise as you wouldn't want to accidentally inhibit it in case a human driver (or just cyclist/pedestrian) also happens to be there and broadly matches the location/angle you expect your other Waymo car to be at.