I run a total same model both in k_eigenvalue and depletion mode, but I note the different values of k, despite the difference is small. The result is below:
k_engenvalue 1.16277
deplete: (without setting the reflector and clad depletable) 1.16330
deplete: (setting the reflector and clad depletable) 1.16275
Thanks for all your advice!
The most likely cause for this discrepancy is a different set of cross section data being loaded at runtime for the depletion simulations (to support all of the transmutation tallies), which changes the ordering/composition of reaction CDFs (and therefore the RNG state / resulting neutron chains). This will get wiped out as you run more particles (the models are the same).
A less likely cause is floating point non-associativity (caused by a different number of threads being run between k-eigenvalue and depletion results) between the timestep 0 depletion results and the standalone k-eigenvalue results (resulting in different neutron chains between the runs). As you increase the number of particles, this will also get wiped out.
I wouldn’t worry too much about this set of discrepancies. They appear to be within a 1 sigma standard deviation of each other (statistically irrelevant). If the differences were outside a 3 sigma standard deviation band of each other I would be far more concerned.
