Status of the Celestial Reference Frame

Since the advent of the ICRF1 in 1998 as the formal realisation of the ICRS, the successive solutions (ICRF2, ICRF3) have been produced from VLBI astrometric observations of distant compact sources in the radio domain. Originally in the S/X-band, this was recently extended with independent solutions in the K- and X/Ka-bands, allowing already to see systematic differences, probably not all linked to the change of the emission centre with the wavelength. This choice of a solution in the radio domain, was not imposed by  ICRS prescriptions, but just a contingent temporary situation when only the VLBI technique was in position to reach a sub-mas astrometric accuracy for this realisation, widely outperforming the available optical astrometry. As expected this has dramatically changed in the recent years with the optical realisations achieved with the Gaia spacecraft released in 2018 and 2021. An IAU resolution in September 2021 has now recognised the availability multiple solutions and redefined the fundamental reference frame accordingly.
I will present in parallel the main properties of these solutions, together with the key issues that have emerged from their comparisons. We see now plain differences that clearly won't be resolved by more accurate solutions and give evidence of positional difference in a large fraction of the common sources, to such an extent that one must select strict procedures to ensure that these frames have compatible fundamental planes and origins.