trim over ZZ[x] returns a strictly smaller ideal

R = ZZ[a,x,y]
I = ideal(a^2+5, (1+a) * x - 2 * y, 3 * x - (1-a) * y)
J = ideal(x * y - x^2 - y^2 - 1)
K = ideal(3,1-a)

trim(trim(I+K)+J) -- right answer
trim(trim(I+J)+K) -- wrong answer
trim(I+J+K) -- same wrong answer

I guess this kind of thing has come up before:
https://github.com/Macaulay2/M2/issues/146