In preparation for the CROSS experiment to search fordouble-beta (2β) decay of 100Mo, we upgraded alow-background facility at the Canfranc underground laboratory(Spain) by installing a magnetic damping-based detector suspensioninside a pulse-tube cryostat. We tested the suspension inlow-temperature (15 mK) measurements with two scintillatingbolometers made of large-volume crystals of 116CdWO4(reference of the set-up) and Li2MoO4 (R&D of the CLYMENEproject in view of the CUPID 2β experiment), each coupled to athin Ge bolometric light detector. Despite the evidence of aresidual pulse-tube-induced noise, picked up by cabling, we achievedhigh performance with all tested devices. In particular, the energyresolution for 2615 keV γ-ray measured with both116CdWO4 and Li2MoO4 bolometers is ∼6 keV FWHM,among the best-reported results for thermal detectors based on suchcompounds representing a great interest to searches for 2βdecay in 116Cd (Q-value is 2813 keV) and 100Mo(3034 keV) respectively. Highly efficient particle identification(α vs. β/γ) is achieved also with thescintillating bolometer made of a low light yield crystal(Li2MoO4), thanks to the low noise (10 eV RMS) exhibited bythe light detector exploiting the Neganov-Trofimov-Luke effect forsignal amplification. We also found a rather high level ofradiopurity in the Li2MoO4 crystal; only traces of 210Poand 226Ra were detected (∼0.1 mBq/kg each), while the228Th activity is expected to be at least an order of magnitudelower, as well as a 40K activity is below 6 mBq/kg.