Actually the strands are not as mechanically separable as they might seem. Most GPL-centric licensing disputes come down to a fundamental conflict on one idea; whether software licenses cross dynamic linking boundaries or not. The FSF believes that they do, because when combined into a process that executes on a host, the whole combined piece of software can be observed to be a functional unit, and it would make no sense to have a license that only covers components of that unit. To keep the rest of the world at bay while maintaining this opinion, the LGPL was created, which (even in FSF's opinion) does not cross dynamic linking boundaries, and allows for static linking as long as the user is allowed to re-link the executable on demand.

The presiding view of the rest of the world is that license violations cannot occur at runtime without a EULA (i.e. a contract) that tells the user "You cannot do this". Instead, the usual view is that a license violation occurs when GPL source code is included in a separate GPL-incompatible distribution and the combined work is built together and distributed.

The difference between GPL violations occurring at the source level or at the binary one is a very fine one, but is the fundamental source of many emotional debates. It would be nice if once and for all we would get a judge's opinion on whether software licenses cross dynamic linking boundaries or not, to clear up whether e.g. dynamically linking a GPL program to Xlib has any license ramifications whatsoever.