Is there an observed (direct not indirect) speciation on record showing a species shift within the span of history of recorded science?
Could not fit in a comment…
What kind of observations will you accept to be an observation? If we can demonstrate showing genetic data that two current species where actually only one some time in the past would it represent an observation to you? Or does it has to be a lab experiment (experimental evolution)? Experimental evolution with big animals take much time, I am afraid we probably have no evidence (not counter-evidence) of a mammal species that evolve to be phenotypically speaking as different than a fruit fly and a june bug (example coming from your comment). Would you accept considering microorganisms (I have no example in head though right now)?
As you noted yourself I think, it is difficult to know how different (phenotypically or genetically?) should two sister populations (populations who derived from a common ancestral population) be so that you would consider that one of these populations at least evolved into another "kind" (compare to the ancestral population).