In the summer of 2011, the first on-sky astrometric commissioning of PRIMA-Astrometry delivered a performance of 3 m″ for a 10 ″ separation on bright objects, orders of magnitude away from its exoplanet requirement of 50 μ″ ~ 20 μ″ on objects as faint as 11 mag ~ 13 mag in K band. This contribution focuses on upgrades and characterizations carried out since then. The astrometric metrology was extended from the Coudé focus of the Auxillary Telescopes to their secondary mirror, in order to reduce the baseline instabilities and improve the astrometric performance. While carrying out this extension, it was realized that the polarization retardance of the star separator derotator had a major impact on both the astrometric metrology and the fringe sensors. A local compensation of this retardance and the operation on a symmetric baseline allowed a new astrometric commissioning. In October 2013, an improved astrometric performance of 160 μ″ was demonstrated, still short of the requirements. Instabilities in the astrometric baseline still appear to be the dominating factor. In preparation to a review held in January 2014, a plan was developed to further improve the astrometric and faint target performance of PRIMA Astrometry. On the astrometric aspect, it involved the extension of the internal longitudinal metrology to primary space, the design and implementation of an external baseline metrology, and the development of an astrometric internal fringes mode. On the faint target aspect, investigations of the performance of the fringe sensor units and the development of an AO system (NAOMI) were in the plan. Following this review, ESO decided to take a proposal to the April 2014 STC that PRIMA be cancelled, and that ESO resources be concentrated on ensuring that Gravity and Matisse are a success. This proposal was recommended by the STC in May 2014, and endorsed by ESO.