Aims. We validate a semi-analytical model for the covariance of
real-space 2-point correlation function of galaxy clusters. Methods.
Using 1000 PINOCCHIO light cones mimicking the expected Euclid sample of
galaxy clusters, we calibrate a simple model to accurately describe the
clustering covariance. Then, we use such a model to quantify the
likelihood analysis response to variations of the covariance, and
investigate the impact of a cosmology-dependent matrix at the level of
statistics expected for the Euclid survey of galaxy clusters. Results.
We find that a Gaussian model with Poissonian shot-noise does not
correctly predict the covariance of the 2-point correlation function of
galaxy clusters. By introducing few additional parameters fitted from
simulations, the proposed model reproduces the numerical covariance with
10 per cent accuracy, with differences of about 5 per cent on the figure
of merit of the cosmological parameters $\Omega_{\rm m}$ and $\sigma_8$.
Also, we find that the cosmology-dependence of the covariance adds
valuable information that is not contained in the mean value,
significantly improving the constraining power of cluster clustering.
Finally, we find that the cosmological figure of merit can be further
improved by taking mass binning into account. Our results have
significant implications for the derivation of cosmological constraints
from the 2-point clustering statistics of the Euclid survey of galaxy
clusters.