I have a question, in a sense, about how asymmetric left and right cosets can be when dealing with an infinite, non-normal subgroup $K$ of a (non-abelian) group $G$. Specifically, my question is whether there can exist a subgroup $K$ of a group $G$ such that for some (note crucially: just one, not all) $a \in G$, we find that $aK \subset Ka$ but $Ka \nsubseteq aK$?
We already know from the basic theory of normal subgroups that this is impossible if $K$ is finite or if it is normal. We further know that if (and only if) $Ka \subseteq aK$ and $Ka^{-1} \subseteq a^{-1}K$, then $aK = Ka$ as well, since in fact $Ka^{-1} \subseteq a^{-1}K \Leftrightarrow aK \subseteq Ka$. Thus, any example of my desired situation would have this property for $a$ but not for $a^{-1}$.
I appreciate any input/references, thanks.