I spent... longer than I should have trying to take advantage of this bug to
recreate a stable version of the stereotypical tall iceberg that the original tweet was complaining about.
I believe it works because the sail consists of negative mass. The overall mass of the boat is therefore lower, but the part of the hull that's under water produces the same amount of bouyancy based on the area of displaced water. Because of the width of the hull, it's a stable equallibrium.
Interestingly, if the sail were to become submerged, it would create negative bouyancy and the boat would sink!
Any convex shape will tend to minimize the forces due to buoyancy and "horizontal" should always go less deep (or equal for a sphere) than "vertical"
It can be in a local minima of buoyancy without being a global minima. Similar to an un-sharpened pencil balanced on it's end.
(there are about 4 completed saw tooths above the screen)