However if I change the second point of the line, the calculation never finishes, and has to be aborted manually. Notice that the only change is a sign in the x-axis of the second point. Geometrically, this is the second tangent to the circle through the first point.
I have no idea why the 2nd solution never finishes. But you could of course reformulate your problem in an old fashioned way (avoiding “modern” constructs") and you get - at least on my system - almost immediately more or less complicated answers:
This is a smart change in variables. I did a similar thing to make it work, I changed the circle with another which I know touches at the same point from geometry considerations.
The question which remains is, does it makes sense to file a bug?