Just to end this thread, as per Ken's suggestion, page 46 [R] anova, is
correct, and corresponds to Model 1 below, not Model 2. The EU (here,
live animal) gets exposed to each treatment level in a repeated
measures ANOVA, which is the case for both models below. I mistakenly
thought that with subsampling, subunits should receive each treatment
as well, but this need not be the case. Only the EU needs exposing to
each treatment level for a nested repeated measures ANOVA. And my
thoughts about nesting were wrong in the prior notes too. Sorry.