Well actually...On the bare bones fresh install, loading nlme gives melibrary(nlme, lib.loc = "/usr/lib/R/library"). That is version 3.1-155 from r-base. I can manually install from RStudio version 3.1-157. That version (157) of nlme loads, no problem.

Side-by-side the contents of usr/lib/R/library are identical between fresh install new machine and old machine running successfully for years. Anything else installed over base R installs into home directory R library. Same for both machines.

The difference is I can see both versions of nlme in the Packages window of RStudio on the new machine. I can run 157 but not 155. I'm locked out of base nlme on the new machine. But if I go into the nlme folder in usr/lib/R/library on the old machine, I can see that it is version 3.1-157. So the old machine can access and update the folders in usr/lib/R/library, but the new machine is locked out.

I can just load those core packages freshly into the home directory to access them I suppose. But you're right there is some kind of path/authorisation problem I have never encountered before...

lib.Paths output is identical on both machines, whether I look via RStudio or raw R in the terminal.

R_LIBS_USER from both machines in etc/R/Renviron are identical (same R version and OS difference is just 21.10 v 21.04)

So this is not a Bioconductor problem clearly now, but any thoughts or advice or resources you could recommend?