I came across this one today.
$$\int_0^1\int_0^1 \frac {x^2-y^2}{(x^2+y^2)^2} dy dx$$
What I saw was that for any $(x,y)$ in $[0,1]\times[0,1], f(x,y) = -f(y,x)$
and since this region is symmetric across the line $y=x,$ this suggests to me that this integral will evaluate to $0$.
I applied a change in coordinates.
$u =x+y, v = x-y\\ 2 (\int_0^1\int_{-v}^{v} \frac{uv}{(u^2+v^2)^2} du dv + \int_1^2\int_{v-2}^{2-v} \frac{uv}{(u^2+v^2)^2} du dv)$
That appears to me to also evaluate to $0$.
But wolfram-$\alpha$ says $\frac {\pi}{4}$
Where is the hole in my logic?
As Frpzzd noted in a comment, each of your "it's zero because of cancellation" symmetry arguments assumes $\infty-\infty=0$. But once you've noted $$\int_0^1\frac{x^2-y^2}{(x^2+y^2)^2}dy=\int_0^1\bigg(\frac{1}{x^2+y^2}-\frac{2y^2}{(x^2+y^2)^2}\bigg)dy=\bigg[\frac{y}{x^2+y^2}\bigg]_0^1=\frac{1}{1+x^2},$$the rest is trivial.