If you run your system with /var/run as tmpfs that directory gets emptied every reboot, resulting in /var/run/samba being removed. Since net-fs/samba in gentoo relies on this directory existing samba then fails to start. So there are two solutions to this, either have samba init script create /var/run/samba on start if it does not exists, or have samba use /var/run directly. I leave the choice up to the maintainer.
You're not supposed to wipe the contents of /var/run; see Bug 177736, for instance.
but then we got bug #332397 and especially the last comment therein. also this feels a bit like a non-issue. adding a "[ -d /var/run/samba ] || mkdir -p /var/run/samba" seems not like something that needs arguing about when some people like me think /var/run should be fully volatile, and some does not depending on how they intepret fhs (which by the way states: "Files under this directory must be cleared (removed or truncated as appropriate) at the beginning of the boot process." so the argument about wiping /var/run is mostly about if this should be interpreted as if the same holds true for directories in /var/run and not only files).
the exact reason for /var/run being nuked doesnt matter too much. people shouldnt have to re-emerge a package just to create a few dirs/files in it.
Now init script creates /var/run/samba if it doesn't exist.