When tasked with reversing the order of integration in the following integral:
$$\int^1_0\int^{y^2}_{-y}x\ \mathrm{d}x\mathrm{d}y$$
It seems sensible to sketch out our situation:
we have that $-y\le x\le y^2$ and that $0 \le y \le 1$ which gives the following enclosed region:

However, where is the fallacy in writing:
$$\int^1_0\int^{y^2}_{-y}x\ \mathrm{d}x\mathrm{d}y = \int^1_0\left[\int^{0.1}_{-y}x\ \mathrm{d}x + \int^{y^2}_{0.1}x\ \mathrm{d}x\right]\mathrm{d}y$$
(Where 0.1 is chosen arbitrarily) which will then give an enlarged region and thus a larger answer:

This seems clearly wrong, but there's something I'm missing when we have both upper and lower limits of an "inner" integral being a function of $y$. If anyone can spot the exact assumption I'm making which is wrong, that would be appreciated.

The outer integral is evaluated for $y \in [0,1]$, and the inner integral's limits are on $x \in [-y, y^2]$, which you partition into $x \in [-y, 0.1] \cup (0.1, y^2]$. So if $y \in [0, \sqrt{0.1})$, the integral over the second interval is negative, because the upper limit of integration is smaller than the lower limit of integration. For example, if $y = 0.2$, then $y^2 = 0.04 < 0.1$, and the two inner integrals in your expression become $$\begin{align} \int_{x=-y}^{0.1} x \, dx + \int_{x=0.1}^{y^2} x \, dx &= \int_{x=-0.2}^{0.1} x \, dx + \int_{x=0.1}^{0.04} x \, dx \\ &= \int_{x=-0.2}^{0.1} x \, dx - \int_{x=0.04}^{0.1} x \, dx \\ &= \int_{x=-0.2}^{0.04} x \, dx. \end{align}$$ This resolves your contradiction.