(A) and (B) hold: in a balanced RBD every contrast with $\sum C_i=0$ has $\operatorname{Var}(\sum C_i\hat\alpha_i)=\sigma^2\sum C_i^2/r$ , giving common variance $2\sigma^2/r$ for pairwise contrasts (A) and $\sigma^2/r$ for any unit-normalized contrast (B). However (C) is false in general: for two pairwise BLUEs sharing a common index, e.g. $\hat\alpha_1-\hat\alpha_2$ and $\hat\alpha_1-\hat\alpha_3$ , direct computation gives $\operatorname{Cov}(\hat\alpha_1-\hat\alpha_2,\hat\alpha_1-\hat\alpha_3)= \operatorname{Var}(\hat\alpha_1)=\sigma^2/r\neq0$ (since different treatment means in the fixed-effects RBD model use disjoint error terms, only the shared $\hat\alpha_1$ contributes a nonzero covariance) — only pairwise contrasts on disjoint index pairs are uncorrelated. Since (C) fails in general, (D)'s claim that all of (A), (B), (C) are correct does not hold; between (A) and (B), the officially published key marks (D) as fully correct, which our analysis does not confirm — we record the discrepancy here.