R spatial follows GDAL and PROJ development describes the main points of the status now. sp uses WKT2 CRS with PROJ6+/GDAL3+. sp and rgdal read, write, and project and transform objects using PROJ strings before PROJ6/GDAL3 and when raster calls
rgdal::rawTransform() with PROJ6+/GDAL3+. sp and rgdal read, write, and project and transform objects using WKT2_2019 strings from PROJ6/GDAL3. The mechanism used is described in the R-spatial blog.
This rolling RFC (not many comments) depicts the state of play from October 2019 to April 2020, involving a lot of reverse dependency testing of almost 1000 CRAN packages. When the development versions of sp (>= 1.4-2) and rgdal (>= 1.5-8) are submitted to CRAN, a fair number of reverse dependencies will be broken (as with recent sf releases). Some maintainers sensibly find that fixing the PROJ6/GDAL3 transition is sufficiently invasive to make it sensible to re-base to sf from sp. Others are regretably unresponsive so far, many find it hard to check on PROJ6+/GDAL3.
For further links (in addition to the blog post), see (https://github.com/r-spatial/discuss/issues/28), and sf github issues: https://github.com/r-spatial/sf/issues/1146, https://github.com/r-spatial/sf/issues/1187, https://github.com/r-spatial/sf/issues/1231, https://github.com/r-spatial/sf/issues/1328 and many others.
We’ve established that we should have preferred WKT over PROJ strings starting eight years ago: https://lists.osgeo.org/pipermail/gdal-dev/2012-November/034558.html (read the last paragraph