In quantum gravity one usually splits the metric $g= \bar{g}+h$ into a background field $\bar{g}$ and a fluctuation field $h$. In order to obtain a propagator one has to gauge fix the action (e.g. Einstein-Hilbert) which breaks the gauge symmetry
$$h\rightarrow h+\mathcal{L}g$$
$$\bar{g}\rightarrow \bar{g}$$
where $\mathcal{L}$ is the Lie derivative.
However, one can choose now the gauge fixing such that it is invariant under
$$h\rightarrow h+\mathcal{L}h$$
$$\bar{g}\rightarrow \bar{g} +\mathcal{L}\bar{g}.$$
My problem is now that I don’t see why for example the de-Donder gauge condition
$$ F_\mu[\bar{g},h]=\bar{\nabla}^\nu h_{\mu\nu}-\frac{1}{2}\bar{\nabla}_\mu h $$
is invariant under the latter transformation.
References:
J. Pawlowski, M. Reichert Quantum gravity: A fluctuating point of view (eq. 46)
N. Christiansen From the Quark-Gluon Plasma to Quantum Gravity (p. 35)