I tended to avoid updating #Rstats because then I had to reinstall a large number of packages (including commercial ones). My perhaps dangerous approach to more frequent updates:
all my R packages live in a folder defined in .Renviron as R_LIBS_USER="~/r-packages". Then I update.packages()
It has never failed though. At the same time, I usually wait a bit before rolling a new version. I tend to be software version conservative.