Researching it some more led me to an article from October 2017: "Waymo says its cars can handle nighttime driving and light rain, but its cars will be geofenced" [0]. That should handle a decent amount of it but I definitely agree that the whole "what to do mid taxi service if conditions become terrible or where to park the fleet" will still be a thorny problems for some time.
That being said, unless you think self driving cars will never be able to handle 99% of situations, those problems will just cause the self driving car revolution to happen more gradually as they slowly adopt more and more unique situations (which seems to be their current strategy starting with phoenix). During that gradual progression, I just see them slowly eating away at ubers/lyfts market as more markets become locally profitable when they can adapt to the local conditions/economics (and it is not like some other companies self driving tech will be instantly be able to handle more complex situations in a local market that the leader of the market cannot handle)
[0]: https://arstechnica.com/cars/2017/10/5-things-we-learned-fro...
Rain and snow are normal conditions in most of the U.S. Here in NYC (one of Uber's largest markets) it's been raining a lot lately. I don't think a taxi service that doesn't operate in the rain would get much business here. The demand for taxis actually goes up here when it rains.
Also, it can start raining quite suddenly. What do you do if someone is on their way to somewhere in a self-driving car and it starts to rain?
> It will almost certainly be profitable to just service customers when weather conditions are fine and then park their cars in bad weather.
You'd need somewhere with enough space to park your entire fleet when it's raining or snowing. That would be rather expensive.