Uncorrelated errors give non-identical, uncorrelated responses

In a multiple linear regression set up $\underline y=X\underline B+ \underline E$ under the assumption $E(\underline E)=0,\ \operatorname{Var}(\underline E)=\sigma^2$. In the response variables $y_1,\dots,y_n$ are: