K-factor > 1 across the band in simulation, but the PA still oscillates at 8GHz in the lab. What do you check beyond K-factor? Mu-factor? Something in the bias network?