InLowestTermsOverTheGaussianRationals​(AngouriMath.​Entity,​AngouriMath.​Entity)

The polynomial gcd reads its coefficients in the rationals, and i is not one, so
a constant with it in was left as the expansion gave it and nothing in it cancelled.
Newton's iteration for the inverse modulo a power of a quadratic squares its iterate
every round, and the split of 2 d^3 t^2/((t^2 - c - i d)^3 (t^2 - c + i d)),
which the root substitution makes of Rubi's 4.3.2.1
(c + d tan(x))^(5/2)/(a + i a tan(x))^2, came out with d^572 in its
coefficients and an answer of 240,000 characters, which the default precision could
not evaluate.
https://github.com/asc-community/AngouriMath/issues/1788
The unit is read as a symbol j with j^2 = -1, so each side is
A + j B with A and B free of it; the denominator times its
conjugate is real, B0^2 + B1^2, and what the three polynomials over the
rationals then share is their gcd. With the denominator real and that gcd one the
form is unique, so a constant that is zero comes out zero.