What does the filesystem hierarchy standard really mean by "local data"?

since isn't the entire filesystem usually local to the host?

No, sharing the read-only parts (e.g. all of /usr or even the entire /) across the network was actually quite common.

Although I think that's not the point. "Specific to this host" can also mean not provided by the distribution – it's software you've built and installed by hand, as opposed to the rest of /usr containing generic packaged files (which would indeed be the same across many hosts).

For example, many source code build systems use /usr/local as the default install location. If you just run ./configure && make && make install, it will dump everything to /usr/local/bin and such.

Expanding on the answer by @grawity:

The term "local" stems from a time, when there was only a single UNIX. In this context, all systems running the same version would have a big set of files identical between all hosts, with accordingly only a minority of other files differing between hosts.

These non-universal files would either be configuration data in /etc, or optitional files in /opt or "something else, local to the host" - i.e. */local.

In modern times, when there are (wildly) different members of the UNIX family tree and storage requirements for the OS being a tiny fraction of the total storage typically available, many coming in different flavours (*BSD, Linux distros), the term "local" has been watered down, maybe even into a historical relict.