Fermat noticed that in the right triangle $(156, 1517, 1525)$ the square of the difference of the legs exceeds the double of the square of the least leg by a square, so being $(a, b, c)$, other than $a^2 + b^2 = c^2$, the three sides of this right triangle also obey,
$$(a - b)^2 - 2b^2 = d^2$$
For those kind of triangles, one can prove that
$$2c^4 - d^4 = g^2\tag1$$
where $g = (a+b)^2 - 2b^2$.
To compare, the quartic Diophantine equation,
$$2z^4-y^4 = x^2\tag2$$
has infinitely many solutions:
$(x, y, z) = (1, 1, 1), (239, 1, 13), (2750257, \color{red}{1343}, \color{blue}{1525}), (3503833734241, 2372159, 2165017), (2543305831910011724639, 9788425919, 42422452969), (76285433470805578504147559981041, \color{red}{5705771236038721}, \color{blue}{7658246457672229}),(\small{21823524993202203117598843430945898516918043105969513349188339015638198841921},\color{red}{75727152767742719949099952561135816319},\color{blue}{126314830357375266295717376544111167953})$
So now is possible to find the sides of those infinite Fermat triangles:
$(a, b, c, d) = (1517, 156, \color{blue}{1525}, \color{red}{1343}), (7518954988589021, 1453978915260300, \color{blue}{7658246457672229}, \color{red}{5705771236038721}), (121931332515412369802346183394249940305, 32987672231129530249809032005100126928, \color{blue}{126314830357375266295717376544111167953}, \color{red}{75727152767742719949099952561135816319})$
I noticed that some third solution (the $z$) of the quartic represent a solution (the $c$) of those Fermat triangles. Why is that?