I think we interpreted this question differently. It looks like you specified T as the weighting matrix, which is what I initially thought as well. If you look at slide 14 in the Multiple Linear Equation Models lecture, it says that $$\beta = E(T'X)^{+}(T'y).$$
So for OLS, I think it should be T=X since the OLS formula for beta is (X'X)^{+}(X'y). Thus, T is random since X is random. (You wrote T is not random since it is a constant)
I think we interpreted this question differently. It looks like you specified T as the weighting matrix, which is what I initially thought as well. If you look at slide 14 in the Multiple Linear Equation Models lecture, it says that $$\beta = E(T'X)^{+}(T'y).$$ So for OLS, I think it should be T=X since the OLS formula for beta is (X'X)^{+}(X'y). Thus, T is random since X is random. (You wrote T is not random since it is a constant)