Why would it turn 'belly up'?
Every sinking is different.
Estonia flooded through the open bow visor, it listed and the machinery spaces flooded through deck and superstructure openings.
Why does it need to turn 'belly up'
That's something you invented. Very few ships sink 'belly up'
Indeed, a ship sinks for exactly and only one reason: it has lost buoyancy. That loss can occur for any of a number of reasons, and with the ship in any orientation. Stated better, the ship's orientation is utterly irrelevant to its buoyancy.
How a ship maintains roll stability is an entirely separate matter of physics involving the center of gravity and the center of buoyancy. There is a simple model for reckoning this in terms of a derived property called metacentric height. The righting moment in a roll is determined by the righting arm, a conceptual geometrical construct in the model. The way the ship behaves under normal circumstances is governed by the simple form of this model, which includes the concept of the angle of vanishing stability, where the direction of the righting arm reverses, the roll rate increases, and the ship capsizes.
What's important about this model is that it deals with
external forces that cause roll, such as wind and waves. Consequently it fixes the center of gravity and allows only the center of buoyancy to vary. This is what is meant when the model is said to work only for an intact hull. If the vessel begins to ship water for any reason, the free surface effect preferentially directs water to the low side of the ship. This then causes the center of gravity to become a variable in the model instead of a fixed point. (The same thing can happen with shifting cargo.) When this happens, the righting arm doesn't work as intended in the intact-hull case. The simple form of the model no longer describes the physical behavior of the vessel. One of the things that can happen is that the righting moment can be zero with the ship at some arbitrary roll angle. Specifically the righting arm has a length of zero with the ship at that roll angle. There is no physical law dictating that a ship taking on water must continue to attempt to right itself, or indeed must increase its roll rate as the former angle of vanishing stability is exceeded. In effect, the computation of that angle is now dynamic owing to the independently varying centers of gravity and buoyancy.
Further, the center of buoyancy is merely the center of mass of the volume of water displaced by the vessel. In the orientations contemplated by the naval architect, this would be restricted to a fairly manageable envelope of values. A foundering ship quickly exceeds that envelope. Again, the external-force physics of ship roll stability break down quite quickly when the vessel begins to take on water.
Finally, as we have discussed at length and then witnessed in the loss of the sailing yacht
Bayesian, the important angle is usually not the angle of vanishing stability, but the downflooding angle. This is the roll angle at which the first opening into the hull passes below the water surface and allows water to begin entering the hull. That's the angle at which center of gravity starts to become a variable rather than a fixed point. And it's also—separately—the time when buoyancy begins to decrease.
No two sinking vessels progress through all the possible physical states, and no two sinking vessels will progress through them in the same order or the same rate. And this part of the physics dialogue is where Vixen utterly flunked the test. She got hung up on the basic concepts of points, lines, and vectors—those things you start learning about in primary school. And she could not contemplate at all what would be needed to adjust the model to account for flooding. So no, when Vixen tries to explain how ship's sink, the premise, "...and I know what I'm talking about," was proven false long ago.