See a palette the way readers with protanopia, deuteranopia, tritanopia or achromatopsia see it, and get told which pairs collapse into each other.
The same picture through each kind of color vision. Green against red is where most of it goes.
The colors you actually ship, run through the same simulation. A pair that ends up the same fill has lost whatever it was signalling.
Each color is converted to linear light and passed through the standard confusion matrix for that deficiency, then converted back. It is the same approach the well-known simulators use.
A pair that is distinguishable normally but falls under a 1.2:1 contrast ratio once simulated. Below that the two fills are effectively the same color, so anything you signalled with them is gone.
Carry the meaning twice. A label, an icon, a pattern or a difference in lightness all survive every deficiency, and lightness in particular is what the simulations here leave intact.