crop or mask large raster on inport

Dear list,

Is there a way to reduce the spatial extent of an impractically-large
raster (~50Gb) on import? Can a crop or mask be applied as the large raster
is read into R to create a more manageable subset?

I've looked at raster and rgdal helpfiles, but have not found this
functionality. This would be akin to the sf::st_read() 's new "query"
argument that takes SQL statements for filtering records while reading data
into R.

The ultimate use case is to summarize a large categorical raster into many
thousands of polygons, and I thought reading small subsets in at a time
could be looped through or parallelized for summarizing smaller pieces of
the data. If there are other approaches to zonal functions (like
equivalents of ArcMap's Tabulate Area) that could also work.

Thanks for your consideration, Dexter

        [[alternative HTML version deleted]]