Three surprising open topics, puzzles, in robust control are highlighted. The first and the second puzzle deal with difficulties in generalizing H-infinity control theory to realistic persistent signal setups, It seems difficult to generalize the theory to such setups and simultaneously keep the terminology robust H-infinity control intact. These puzzles have implications also to control-oriented model validation. The third puzzle, due to Georgiou and Smith, deals with difficulties in the doubly infinite time axis formulations of input-output stabilization theory.