Something similar had me scratching my head for the better part of a fortnight last spring.
I'm running two Victron 100/50 MPPTs off a mix of panels — nothing exotic, just some Renogy 200W units and a couple of older 300W panels I picked up second-hand — feeding into a 300Ah Fogstar Drift LiFePO4 bank with a SmartShunt keeping tabs on everything.
What I noticed was the SmartShunt's state of charge drifting further and further from what the MPPTs were reporting in VictronConnect. Some days it was a 12% discrepancy by mid-afternoon. Not catastrophic, but deeply unsettling when you're trying to trust your system.
Turns out the culprit was the synchronisation between the two controllers — they weren't networked together via VE.Smart, so they were essentially making independent decisions about absorption and float. One would drop to float while the other was still pushing, and the shunt was getting confused about what "full" actually looked like.
Enabling VE.Smart networking through the Victron app so both MPPTs could see each other's data sorted it almost immediately. The charged voltage threshold on the shunt also needed nudging to match what the batteries actually needed rather than the default.
Worth asking — are your controllers talking to each other, or are they operating blind? And have you double-checked your tail current setting on the shunt? That's often the quiet troublemaker nobody mentions.
Curious whether others have seen similar drift issues with parallel controllers on a shared bank. Feels like this comes up more than the manuals suggest it should.