What about a design change: connect the closed end of the Pitot tube to a high pressure pump. Whenever erratic air speed indications are suspected, push a burst of fluid via this connection to dislodge ice, dirt or leftover scotch tape.
This gives me another idea... How about reversing the airflow completely, ie., push compressed air out of the sensor chamber continuously and measure the difference to the expected in-chamber pressure.