3 different IMU's raw data locking up a Mega (NEW INFO)

NEW INFO.
I have installed the LED13 into the POLOLU loop code to see if the program was stopping when the screen freezers.
The LED stop's flashing when screen live data stops, other times the screen stops, LED still blinking.
Also the PC looses com's to the mega port.
Some times when reconnecting the serial lead(to try to reconnect serial), the LED will be blinking (when power reconnected program running) other times it won't and l have to upload code into mega to get serial back(program wiped).

Mega green ON LED lit at all times.

Hi everyone, l have a Mega board that is connected to my laptop via the serial cable using Ardunio 1.6.9.(also same problem with older versions).
I have tried a BNO055(raw data example), HMC5883L (calibration +heading example) and a
Pololu IMUMINV3 (heading +Calibration example) all of which stop sending data to the serial monitor. The monitor comms is still live (with dead/frozen data) as the lines keep scrolling sometimes and other times it stops scrolling.
The Pololu IMU is soldered to the back of the mega, the others are via 300mm male female cables to the other Megas.
I have tried 3 different Megas, all with the same result.
The Pololu unit has had the Accel stop, then later the Mag stops sending(in heading example, the gyro isn't used).
The time frame can be anything from < 5 seconds, quite often 20 second and up to 2-5 minutes.
The programs are all examples from the IMU libraries (l have deleted the libraries and reinstalled from zips), so nothing l have altered.
The Mega's are 2 original and 1 Chinese copy(Duinotech).
I have tried powering the Mega with 12-4.95V via the 5V rail, 12V to the VIN/plug and a Auto pilot code (that works correctly, confirmed by the programmer) with LCD, so l could run without the serial monitor/PC cable (powered via 5V or 12V as above) and that also locks up.
The voltage on the 5V rail with the serial cable only connected, via a DVM is 4.965V.
The same connection as above gives 4.4V on the VIN, (so this after it goes reverse thru the Volt reg?)
I have supported the wires when moving the IMU around and checked for losse connections, without seeing a problem.
I have tried to compile Ardunio CrashMonitor to try some debugging but can't get it to compile, so can't currently use that(but there should be nothing wrong with the example code).
It would seem to be a hardware/wiring issue, bu there is little to do wrong. Serial cable, 4 wires to the IMU, 5V +-, SCL + SDA
Any suggestions, keeping in mind that l am basic/intermediate Ardunio user, so have limited knowledge.