I suspect that spice does not handle them as would a code compiler in the same manner.
But the basic operation is the same in the long run.. With LIBS, you pick out what you need, INCLUDE, you bring it all in except the comments and maybe even convert it to some sort of internal token language. The extraction of selected pieces of subckts from LIBs can also be condensed down into the same token type stream of data..
Since LIBS apparently do no have an fast search index in the LIB file, due to its nature of being a Text file, that leads me to believe that at some point Ltspice needs to scan it all and build an index..
THat's just my beef!