What dictates the set of microcanonical ensembles of a system in the canonical ensemble? So in the canonical ensemble we have a system inside of a heat reservoir, which can exchange energy with the system. This means the energy of the system is not necessarily a constant value but determined effectively by the local energy distribution of the reservoir.
What I'm essentially asking is if we took two "snapshots" of the canonical ensemble at two different times and looked at the energy of the system would we have some probability of finding the system to be at some energies E1 and E2 where they differ only because the system happens to be in a "colder" or "hotter" region of the reservoir due to the random distribution of energy throughout the reservoir?
 A: Equilibrium statistical mechanics deals with systems in a canonical state in which, by definition, the temperature is constant. There are no temperature fluctuations by definition.
The canonical ensemble is a hypothetical collection of an infinite number of non-interacting systems having fixed T, V, and composition. Even if the members of the ensemble are macroscopically identical, they show a wide variety of microstates, since many different microstates are compatible with a given macrostate.
The possible microstates are found by solving the Schrödinger equation
$\mathscr{H}\psi_j = E_j \psi_j$
The possible wave functions \psi_j and quantum energies E_j will depend on the composition of the system
$E_j = E_j(V,N_A, N_b,N_c,\ldots)$
where the system’s composition is specified by giving $N_A, N_B, N_C$, the number of particles for each species (for example consider the particle in the box model in which the translational energy of a particle in a cubic box depends on the volume of the box).
As you can see from the previous equation $E_j$ does not depend on the system’s temperature. Temperature is a macroscopic non-mechanical property and does not appear in the quantum-mechanical Hamiltonian of the system.
A: What you are measuring is the energy of a single system going in time through its canonical micro states. As time goes you are measuring the time variation of the system property not the ensemble one.According to the Ergodic theory the time average of the system energy (over a wide interval of time) exist uniquely and coincide with the ensemble  average and this is the utility of introducing the hypothetical ensemble in the first place .In Equilibrium (or steady state) statistical mechanics the ensemble and its systems are in the  time independent steady state. If you are getting two snap shots (meaning the measurement is carried over zero interval of time ) for the system one micro state at the same point in space you should get the same value of the energy density and if the two  points differ in space, the readings could be different due to different spatial  distribution of the energy of the reservoir.
One may add that your system can be categorized as a steady state statistical mechanics and not equilibrium statistical mechanics.
A: 
So in the canonical ensemble we have a system inside of a heat reservoir, which can exchange energy with the system. This means the energy of the system is not necessarily a constant value but determined effectively by the local energy distribution of the reservoir.

I would say that "this means the energy of the system is not fixed but it varies according to the exchanges of energy between system and thermostat.". It is not the local energy distribution  in the reservoir which matters but the amount of exchanged energy. However this is a side remark, although important to recast in a more correct way your guess/question.
The answer to you question is affirmative. The probability distribution in energy (of the system) in the canonical ensemble is given by 
$$
P(E) = \frac{g(E)e^{-\beta E}}{Z}
$$
where the normalization  $Z=\int P(E) dE$ (the partition function) depends on $\beta$ but not on $E$. The degeneracy factor $g(E)$ is the density of states in energy, measuring how many states are in an interval of energy centered at $E$.
Therefore, two microstates of the canonical ensemble of energies $E_1$ and $E_2$ will have probabilities in the ratio $\frac{g(E_1)}{g(E_2)}e^{-\beta (E_1-E_2)}$.
Just a final comment to Ismail Abbas' answer, all that is pure equilibrium statistical mechanics.
