FD solver not always calling wake/0 when needed

Dammit, I was trying to find these kinds of bugs in *my* solver, by
comparing it to FD, not the other way around...

I don't really plan to mess with FD to fix this (a quick look suggests
I'd need to think too hard).

Note the value of the upper bound of Z in each case.

[eclipse 8]: [X, Y] :: -10..10, X + Z #=< 0, 3 * X + 2 * Y #>= 0.

Z = Z{[-10000000..10]}
X = X{[-6..10]}
Y = Y{[-10..10]}

Delayed goals:
        0 - X{[-6..10]} - Z{[-10000000..10]}#>=0
        0 + 2 * Y{[-10..10]} + 3 * X{[-6..10]}#>=0
yes.
[eclipse 9]: [X, Y] :: -10..10, X + Z #=< 0, 3 * X + 2 * Y #>= 0, wake.

Z = Z{[-10000000..6]}
X = X{[-6..10]}
Y = Y{[-10..10]}

Delayed goals:
        0 + 2 * Y{[-10..10]} + 3 * X{[-6..10]}#>=0
        0 - X{[-6..10]} - Z{[-10000000..6]}#>=0
yes.
[eclipse 10]: