I think the directory is created by enewuser pcscd -1 -1 /var/run/pcscd pcscd,openct (also in newer ebuilds). This conflicts with the policy of only creating stuff in /var/run on-demand.
It's not "create on demand" but it's rather "make sure it's there". /var/run/pcscd is created when starting the init script that's what's important (it won't fail if it's not there), but of course the ebuild tries to create it the soonest.
From bug #332633 (the one this bug blocks). --- As discussed on the gentoo-dev mailing list, some ebuilds choose to create a /var/run/${PN} directory at installation time instead of at runtime. Our init.d scripts should check whether a subdirectory is present and create it if need be, instead of using keepdir in the ebuilds. --- This is pretty much contrary to what you wrote. There is no point to creating something in /var/run at build-time, since /var/run can (and will - see bug #361349) be a tmpfs. Now, doing both (ebuild and init.d) might be ok, but I think that writing "of course the ebuild tries to create it the soonest" hints that you might not be aware of bug #332633.
Okay, see, I'm the current QA lead (until proven otherwise) and I've been working _quite_ a bit about /var/run on tmpfs as I started using it way before /run became cool, so I know more or less everything about this kind of issues. In this case, whether it's tmpfs or not, pcscd user should _really_ have /var/run/pcscd as home. And enewuser by default creates the directory in "${D}" if it's not there already. So that's where my "of course" comes from. Since this is still not an issue (okay we create a directory that will be removed at boot, big deal... it'll be created by the init script anyway), this bug can stay closed, okay?
Sure, maybe you can also add a clarification to bug #332633? I.e., about "remove explicit keepdir" vs. "don't try hard to keep /var/run clean". In the bunch of bugs I opened today about /var/run, one was about depending on directory in /var/run, some were about explicit keepdir, some about enewuser, some about directory created by make. Bug #332633 doesn't make enough distinction. The reason I opened those bugs is that I regularly build Gentoo images, so I see the difference between directories created by ebuilds and directories created by services (init.d scripts or binaries themselves), by simply doing ls -l /var/run. So I thought I would contribute to /run migration (which is happening, cool or not).