I made an off-hand remark in the last class (recorded Friday) that while uniform-cost search is guaranteed to be complete and
optimal as long as the action costs are positive (strictly greater than zero), the efficiency itself does depend on the action cost
variance (e.g. if one action costs 0.0001 and another costs 10 units). I mentioned a recent paper by Will Cushing in this context.
If you are curious, you can find that paper at http://rakaposhi.eas.asu.edu/cse471/surrogate-search.pdf