Regarding the proof of the fact that an arrow in a category can have at most one inverse:
Suppose $f:A\to B$ is an arrow with two inverses $g,h:B\to A$. So we have
- $h\circ f=1_A$
- $f\circ h=1_B$
- $g\circ f=1_A$
- $f\circ g=1_B$
I think the proof only uses $(1)$ and $(4)$:
$$h=h\circ 1_B=h\circ f\circ h=1_A\circ g=g.$$
Are conditions $(2),(3)$ really unnecessary, or am I missing something?
They are really unnecessary. Any morphism with both a left and a right inverse has an inverse, which is equal to both the given left and the right inverse. In fact it is more convenient in certain foundational systems like homotopy type theory to define an isomorphism as a morphism with both a left and a right inverse.