Snowflake should self-diagnose where it fails in the connection process, and inform Tor

We have periodic reports (e.g. #40044 (closed)) of people in China saying that Tor Browser + Snowflake gets to 10% bootstrapped and can't get past it. We got another one on irc just now. Our own internal tests show that Snowflake bootstraps successfully on the VPS we're trying it from, but clearly that's not the end of the story. For example, I bet the mobile carriers have different constraints.

I was first thinking to suggest some standalone Snowflake debugging tool that would try a bunch of things and see how they go and summarize it for the user.

But then I realized: Snowflake itself should do this, because it does try things, and it learns how they go, and our users already have it. So all that remains is (a) figuring out which conclusions are worth escalating to the user, possibly including some refactoring inside Snowflake to do the steps in a way where we're confident in our results, and then (b) deciding what the right pathway is for escalating the information.

For 'b', we should be careful to avoid getting bogged down picking between options, since there are plenty of approaches that will do adequately. Maybe the PT log command is useful here, and (if I understand it correctly) in that case the way users can see the output is by preferences->tor->view logs.

And then I imagine the bulk of the work will be in step 'a'.

To get us started: what is the taxonomy of ways that Snowflake can fail to make its connection? And for each of those ways, is there an obvious point where Snowflake can self-assess that it has failed?