Hello Radek,
1) Yes, I meant S12ZVMC. Sorry for the typo!
2) We are using our own board, we are not using any evaluation board from NXP.
3) The output data stream is generated by our application, however we are using the Term+Inhr components of PE from Codewarrior, so essentially it is not our code, we are only calling the functions.
4) Unfortunately we cannot check using oscilloscope or logic analyser because this problem only happens in the field and not in the lab environment. Typically this might get triggered because of some event such as unshielded wires (running from device to PC) is shaken or something. The problem is that even after the wire is corrected the data does not correct itself, unless the device is power cycled. Even restarting the PC software does not correct the problem, so it's not on the PC side.
5) I am not sure if this is the lines are broken in exactly the same place but it could be. They seem to be approximately of similar length to the eye visually
6) The reason we would like to reset SCI is so that when the product is in the field and remains turned on for a long period of time and then if we need to communicate but find that the output is corrupted, at least by reset-ing SCI we could bring the system back into communicable state.
7) We are connecting TX, RX, GND via wires to a UART-USB 'dongle'. Reseting the dongle (remove and reinsert), restarting the PC terminal software etc. does not solve the problem. We have to reboot the device.