Unexpected SFFE behaviour

[trathborne]

Describe the bug
I am trying to write Meltybrot in SFFE for use with XaoS. It looks like this:
https://www.cesoid.com/fractal#al=meltybrot

The algorithm is hidden in minified JavaScript there, but here it is in MathMap syntax:

Ax = 0;
Ay = 0;
Ax2 = 0;
Ay2 = 0;

while ((Ax2+Ay2) < 4) && (a_count < a_max) do
  Ay = 2*Ax*Ay + if a_count % 2 then x else y end;
  Ax = Ax2-Ay2 + x;
  Ax2 = Ax*Ax;
  Ay2 = Ay*Ay;
  a_count = a_count + 1;
end;

In the two-real-number process of computing z²+c, y and x are swapped on alternate iterations via if a_count % 2 then x else y end;

This is a cool enough fractal that I will probably just try to add it natively, but I'm trying to figure out if I have misunderstood something about SFFE syntax.

To Reproduce
This renders a plain old Mandelbrot set as expected:
(powi(re(z),2)-powi(im(z),2)+re(c))+(2*re(z)*im(z)+im(c))i

This is close to, but not quite the Meltybrot rendered by the website or MathMap:
(powi(re(z),2)-powi(im(z),2)+re(c))+(2*re(z)*im(z)+(re(c)*((cos(n*3.14159265358979)+1)/2))+(im(c)*((cos((n+1)*3.141592653589793)+1)/2)))i
To alternately swap the final term im(c) I'm using cos() as a proxy for mod() because SFFE doesn't have mod() or if(). Logic: (cos(even*pi)+1)/2 == 1 and (cos(odd*pi)+1)/2 == 0

Expected behavior
I expected the picture you'll see at the website, but got something slightly weirder.

Screenshots
Web:

XaoS:

Desktop (please complete the following information):

• OS: Linux
• Latest XaoS from git