We have conducted a survey aimed at identifying a sample of z˜7-10 candidates accessible to detailed spectroscopic studies. The deep survey on A2667 covers a large field of view with HAWK-I and FORS2 at ESO/VLT between 0.8 and 2.5μm (typical depth m(AB)=27.5 at 3σ). 13 sources were identified based on the Lyman Break technique, namely 1 J-drop, 8 Y-drops and 4 z-drops over the ˜45 arcmin^2 field of view. Optical and near-IR data were combined with Spitzer and Hubble Space Telescope data, when available, in order to determine photometric redshifts. Although the best-fit is always obtained for a high-z solution, this sample shows a relatively high contamination level from extreme mid-z interlopers which may also be present in other surveys. This is demonstrated by the recent detection of two candidates by Herschel and LABOCA, making the high-z identification unlikely. Waiting for a complete spectroscopic follow up, we have used different criteria to account for this contamination and to derive new constraints on the bright end of the UV LF from z˜7 to 9, using a MC approach based on the redshift probability distributions. Our results are consistent with a significant evolution in the density of bright galaxies beyond z˜8, whereas no clear evolution is observed between z˜7 and 8.