This is a general concern looking for views from those who have upgraded to 3.6 on a Mac (Mac OS Mojave).
Now I have been through real hell when 3.5 came because I had to not just reinstall all the packages on my Mac it was needed on the rstudio server on my AWS Linux server as it soon went out of sync. Which resulted into the need of re installing all packages on the server too.
I see on this thread a similar experience on 3.6.
I guess if it becomes difficult to upgrade R versions in a smooth way we will lose the advantage we have on Python. This was one of the killer advantages I used to tell everyone.
I didn't have the same problem with Mojave, perhaps because I changed permissions in the framework library for the {base} packages and use the personal library for everything else. I don't sync with a server, so I didn't have any issues there.
An important question is: do you need 3.6 ersion of R?
If the answer i not, do not upgrade, no matter of easy or hard it may be...
I hate all this new way of thinking that we need to have the last and closer version of everything, call it daily builds etc... I is jut an insane new way of consumism....
Not o mention how bad is mac when it cames into updating software
I understand your concern — it's no fun to get stuck in dependency hell, and it can really make you leery of upgrading anything ever again.
I'm not sure I have a clear picture of your specific setup, but a couple of thoughts:
I don't know exactly what was wrong in the other linked discussion, but as far as I can tell the root of the problem had to do with system dependencies external to R (a C++ library), possibly related to the operating system version being a few years old. Unless you share the exact same system particulars as that person, I'm not sure it's a representative example of what your experience upgrading to R 3.6 will be like.
R 3.5 was a somewhat unusual case, since there were much bigger internal changes than are typical for that level of R release. However, it has long been the case that packages need to be updated — which amounts to re-installation — for every new minor R release (R's versioning system is major.minor.patch, but you may see people refer to the annual "minor" releases as "major" since the actual "major" releases happen extremely infrequently).
The R for Windows FAQ outlines the traditional practice using update.packages(checkBuilt = TRUE, ask = FALSE).
If you're dealing with anything more complex than a single-user setup and/or you're concerned about reproducibility of analyses or other data products, I really like the framework for thinking about your options outlined in this webinar: