Why does the brand teal shift so much between simulations?
Because teal sits between green and blue on the spectrum, so its result depends heavily on which cone is missing: losing the L or M cone (protanopia, deuteranopia) pushes teal toward blue-purple, while losing the S cone (tritanopia) barely moves it, since tritanopia mainly affects the blue-yellow axis and teal keeps enough green to anchor it.
Do I need to change the brand color if it shifts a lot under simulation?
Not necessarily. An identity color can shift under simulation and still work fine, as long as it is never the only signal for a functional product state; a logo can read as teal to everyone and still be recognized as the brand, distinguishability only really matters where color carries information.
Are dark colors and near-white always safe against every simulation?
They tend to be more resistant, because luminance difference survives better than hue difference alone across all three conditions, but "more resistant" is not "immune": two very dark colors with different hues can still collapse together, so they still need testing, just with lower priority than saturated midtones.
Does this simulator replace a WCAG contrast check before shipping the brand in a product?
No. Hue distinguishability under simulated color blindness and the WCAG 1.4.11 non-text contrast ratio are different checks: a palette can pass one and fail the other, so reviewing both before applying brand colors to an interactive UI is still necessary.