Consider the presentation defined, for an integer $n > 1$, by $$G_{n} = \langle x, y \mid xy^n = y^{n+1}x, yx^{n+1} = x^{n}y\rangle .$$ The group defined by this presentation is trivial. Is it true that the minimum number of cosets that must be defined by the Todd-Coxeter process before any coincidences occur, when run on this presentation over the trivial subgroup, is bounded below by some (at least linear, probably worse) function of $n$, and is such a lower bound known? (In practice, this seems to behave quite badly.)
(A special case of this appears in a recent question, which is what prompted this one.)