Not a great error message for the situation, but I don't think y'[t] == -{1, 0} * y[t], which evaluates to y'[t] == {-y[t], 0}, can be satisfied by any y[t] because the dimensions are different. So there is an error, and one should get an error message.

That said, a message like NDSolve::ndfdmc (about dimensions) would seem more appropriate. The source of the problem of getting the right message might be that at t == 0, NDSolve[] expects y'[0] to have the same dimensions as y[0] == {1, 1}; instead, the ODE sets y'[0] == {-y[0], 0} == {{-1, -1}, 0}; and finally, Dimensions[{1, 1}] and Dimensions[{{-1, -1}, 0}] are both {2} — that is, the computed dimensions agree even though the expressions are not compatible.