In $X = \mathbb R^n$ (finite $n$ to make it simple), we can use either the open balls with the Euclidean metric, $B_d(x, \epsilon)$, or the Box topology (i.e., $n$-dimensional rectangular prisms with finite or infinite sides), as the basis for a topology $\mathcal T$, and these generate the same topology.
We can see this because for any open set $U \in \mathcal T$, for any $x \in X$, we can find a basis element $B$ (from either of these) such that $x \in B$ and $B \subset U$. This picture from Munkres illustrates it well:

This makes enough sense to me. But I can't shake the feeling that they're still different, because they have different "shapes" (i.e., circles vs rectangles).
For example, another way to generate a topology from a basis is by taking the collection of all unions of the basis elements. This paints a different picture of generating the topology, where for any open set of the topology, we must be able to form it by taking a (possibly infinite) union of basis elements.
However, certain open sets of the topology will be formed trivially, or with the union of very few basis elements from one basis, while it may take the union of infinite basis elements from another basis. For example, to produce a rectangle with the "circles" basis, we would need infinite circles of decreasing diameter to get the corner or sides:

But obviously only one from the "rectangles" basis. Is there any significance to this? or, is the significance that in topology, this really doesn't matter, and that's what's important?