I would assume that it is due to the order in which the services are being started. i'd expect that the message bus would start a lot earlier on than slapd, and possibly the same for named. am i right in assuming that if you restart named once you're up and running that it goes just fine? you could easily push slapd further up the rungs in terms of boot priority, but that could theoretically lead into interdependence issues. but essentially the subsystem is trying to reach an ldap server which is not yet running, unlike the clients which obviously just hit the ldap server fine each time.
I'm not aware of any way to fine tune this behaviour to be honest. if you can't move the slapd service up, i'm not sure what your options are - replacing the nsswitch.conf file automatically once slapd starts? but then, servers aren't designed to boot fast, they;re designed to start up once and stay started!