Now I can't believe that we are all supposed to run "make release" on
   machines in 'insecure' mode or in single-user, so I'll assume this is
   a bug?

it isn't so much a bug as a limitation of installing bootblocks on
netbsd/sparc, be it a real disk or a fake disk...i guess the only
real way to fix it is to have a way to inject to the kernel the 
contents of the bootblock, and let it write the sectors itself.