We consider the graviton propagator in a de Sitter background. The propagator depends upon the choice of a gauge-fixing term ${L}_{\mathrm{gauge}=1/2{F}^{2}}$, and we consider the ``\ensuremath{\epsilon} gauges'' with ${F}^{v}$=${\ensuremath{\nabla}}_{u}$(${h}^{\mathrm{uv}}$-\ensuremath{\epsilon}${g}^{\mathrm{uv}}$${h}^{\ensuremath{\sigma}}$${}_{\ensuremath{\sigma}}$). We show that the propagator is completely finite and has no infrared divergences provided that \ensuremath{\epsilon} is not given certain ``exceptional'' values. It is only for these ``exceptional'' values of \ensuremath{\epsilon} that the propagator has an infrared divergence. We then show that in these exceptional cases the divergences are gauge artifacts and are not physical: they make no contribution to any physical tree-level scattering amplitudes. Furthermore, we show that at one-loop order the zero modes which arise (only) if \ensuremath{\epsilon} is given one of the exceptional values are canceled by the Faddeev-Popov ghosts. There is thus no evidence that the de Sitter background is inconsistent when gravitational fluctuations are considered.