The tidal deformability probability distribution extracted from GW170817 alone, or including multi-messenger information, is confronted to astrophysical and nuclear physics additional constraints within a semi-agnostic approach for the dense matter equation of state. We use Bayesian statistics to combine together low density nuclear physics data, such as the ab-initio predictions based on χEFT interactions or the isoscalar giant monopole resonance, and astrophysical constraints from neutron stars, such as the maximum mass of neutron stars or the probability density function of the tidal deformability Λ obtained from the GW170817 event. The posteriors probability distribution functions are marginalized over several nuclear empirical parameters (Lsym, Ksym, Qsat and Qsym), as well as over observational quantities such as the 1.4M radius R1.4 and the pressure at twice the saturation density P (2nsat). The correlations between Lsym and Ksym and between Ksat and Qsat are also further analyzed. Tension is found between the posteriors: the first one is localized in the tidal deformability probability distribution itself, depending whether multi-messenger analysis is included or not, and the second one is between the observational data and the nuclear physics inputs. These tensions impact the predictions for Lsym, Ksym and R1.4 with centroids which differ by 2-3σ. Implications for the nuclear equation of state are also discussed.