I have a PCA9518 in a 3.3V system with 1.24K pull-up resistors. I'm seeing runt clocks on the SCL0 port as though the PCA9518 is incorrectly back-driving the clock from the SCL0 port. The behavior stops if I add 100 Ohm resistors between the microprocessor and the PCA9518 SCL0 and SDA0 ports. Not every clock cycle experiences the runt but, when it occurs, the I2C transaction is disrupted. This is especially a problem with writes to any I2C device, in my case I have EEPROM on the I2C bus.