An exercise from a book that I am currently studying asks to show the following.
Let $B$ be a block of imprimitivity of $\rm{Aut}(G)$ for a vertex-transitive graph $G.$ Then the graph induced by $B$ is regular.
Am I missing something or is a more stronger claim than this true? Namely, that the induced subgraph is vertex-transitive? For if $u,v \in B$ then there is a $g \in \rm{Aut}(G)$ such that $u^g = v$ and clearly $g \in \rm{Aut}(G[B])$ from the fact that $B$ is a block of imprimitivity. Am I missing something here?
Another question that I would like to consider here is :
What are some nice blocks of imprimitivity for some graphs? Are there any nice applications for them?
The only non-trivial blocks of imprimitivity that I could think of are the antipodal vertices of the odd cycle $C_{2n}$ and Hypercube $Q_n$ and a bipartition of a bipartite vertex-transitive graph.
You are not missing anything, the subgraph induced by a block of imprimitivity is a vertex transitive graph.
Constructing blocks: If $u\in V(G)$ then the set of points fixed by $\mathrm{Aut}(G)_u$ is guaranteed to be a block of imprimitivity. The block in even cycles and in the hypercubes arise in this way. If $N$ is an intransitive normal subgroup of $\mathrm{Aut}(G)$, then its orbits are blocks; for example vertex-transitive bipartite graphs. The centre of the automorphism group will often work, provided it's non-trivial. If $G$ is a Cayley graph for a group $\Gamma$ and $\mathrm{Aut}(G)=\Gamma$, then any non-identity proper subgroup gives a non-trivial block of imprimitivity.
Another way of thinking about this is that systems of imprimitivity are just equivalence relations on $V(G)$ that are invariant under $\mathrm{Aut}(G)$. For examples, a graph of diameter $d$ is antipodal if "equals or is at distance $d$ from" is a nontrivial equivalence relation. Hypercubes are antipodal, as is the line graph of the Petersen graph ($d=3$).
If a distance-transitive graph has a block of imprimitivity, the graph is antipodal or bipartite.