The analysis of the DECam COSMOS data has been progressing well. The confluence has been continually updated: https://confluence.lsstcorp.org/display/SQRE/COSMOS+DECam+data+reduction
Some conclusions of the analysis so far:
1) The PSF photometric scatter using the photometric zero points determined the SDSS photometric reference catalog is rather poor, ~2-5% depending on the band.
2) If an ubercal program (which I wrote from scratch for this purpose) is used to determine reliable relative photometric zero points, then the precision increases to ~1% for many/all bands. The difference in PSF and aperture (see #3) photometry likely indicates that there are residuals systematic errors in the PSF models. This is still being investigated. I am also investigating if the PSFEx gives better results than the current PCA code.
3) If the ubercal zero points are used with aperture photometry, then the precision increases further to 0.4% in the g-band and there is no sign of it hitting a floor at the bright magnitudes. This means that with ubercal and aperture photometric we can reach the requirements on photometric precision/repeatability.
4) The astrometric scatter is quite high, ~50-100mas. This is also seen in the validate_drp code. It's not clear why this is happening and will be further investigated.