I agree. The toggle link breech is inherently unstable which is why the thing was used in delayed blowback actions. If you get the thing out of line, it opens!
The toggle actions used in Winchester rifles, if you look at the things you see load being carried by pins. Whatever is weakest in the locking mechanism, that is what is going to fail first. Given the slack in the components and all that, I don't consider the Winchester toggle action rifles all that strong and historically, they were not.
Toggles can be extremely stable, go over top dead center with a buttress on the links to prevent over-rotation, they aren't unlocking, no matter the load. If you look at any toggle lock design the toggle rotates to just over dead center and then stops, careful examination will reveal that they are actually .5 to 1 degree over center. The toggle will not break until the joint is rotated a degree or two.
The "weakness" of a toggle lock is the fact that there are three rotating joints, and since they are rotating joints the have three point contacts, each taking the entire load. Point contacts under high loads leads to very high stresses on the material.
The same "weakness" is also present in any of the "flapper" locks, like the Goryunov, BAR or MAG-58, only they have one rotating joint in these cases.
Shearing a pin (if it is designed properly) is extremely unlikely; however, loosening up due to elongation of the holes is a distinct possibility.
The delayed blow-back of the Pedersen was an interesting application of a toggle that never was positioned in the dead-center (or over dead center), but always slightly broken. The trick was the two links of the toggle were not pinned together, but contacted each other on a cam surface, and the shape of the cam surface was such that as the toggle broke, the contact surfaces remained very close to the axis of the bore so the moment trying to open the toggle was always very small.
The other major delayed blow-back toggle design was the Schwarzlose worked slightly differently, the toggle was not open and lying flat, but closed with both arms superimposed and nearly horizontal, upon firing, the toggle had to unfold, and at full bolt movement was open and flat. Having the toggle closed and horizontal achieved the same purpose as the complicated cam surfaces used by Pedersen, it holds the moment trying to open the toggle small, until the pressure has dropped.