Why is it that /etc/named.reload doesn't appear to cause named to reload
       its database?  (/etc/named.restart certainly makes it do that. :-)
    i don't have either of those files.  i do, however, have a
    /var/run/named.pid which i could easily write those of those
    scripts with.

Ah, I'd typed those scripts in from O'Reilly's DNS book so long ago that I'd
forgotten they weren't standard. :-)  /etc/named.reload is just
"kill -HUP `cat /var/run/named.pid`", which *doesn't* cause named to
re-read /etc/named.boot or to reload its tables.  (/etc/named.restart kills
it dead and restarts it.)

    are you running some other named besides the distributed one ?

Nope, netbsd-current named, built March 4 from presumably fresh sources.
Looking through said fresh sources, I have found what the problem is:
the signal sets a flag indicating that a reload is in order, but it won't
happen until after the next query has been completed.  Never mind.