On the randomised stability constant for inverse problems