Voltage AFE with fast rising edge problem

[Mark]

Hello, 

I am having a problem with an AFE (attached) that uses an INA826 instrumentation amplifier to drive an ADS1271 ADC. 
This is a single 5V supply circuit with a 2.5V reference.  The entire AFE channel is powered by an isolating DC-DC and the digital signals also go across the isolation barrier. 

There is a voltage divider formed from two legs, each 100k and a divider resistor of 806 ohms to give a full scale output of +/-1.7V (referenced to the 2.5V mid point) for a 300VRMS input.  The ADC uses a 1.7V reference.  The measured voltage input can be 300VRMS (60Hz + harmonics) and can come from anything really. 

With sine waves, everything looks fine at the output, but with fast rising edges it falls apart, e.g. a differential square wave starts to show an anomaly at the zero crossing. 

I added some unbalanced stray capacitance between the inputs to the circuit ground and now the simulation shows the same waveforms as I'm seeing on the PCB (so I think I'm in the right area). 
The two x 475k resistors are intended to keep the common mode of the inputs at mid-supply.  These should also provide a return path for the input bias currents for the INA826? 

Any ideas or suggestions welcome, thank you.