The diagnostics Tag shows Good connection without errors. But I have investigated it further and it seems that the poll requests are being sent to all 4 slave devices at once resulting in Frame and CRC errors by slaves.
I think I am experiencing the same problem here: