ElPaso:
Concerning the watchdog. As far as i understand it will only help me in case the sketch hangs in general. Because i can only see the hanging point from disassembling the compiled sketch on my harddrive.
I don't know what you mean by "hangs in general". You do certainly need to wait for a hang to cause the watchdog timer to time out. After that you need to read the EEPROM address you set via WatchdogLog.begin(), then you need to disassemble the compiled sketch and find the address at which the watchdog reset occurred. If the results are not conclusive t might be helpful to repeat this process multiple times to see whether the results differ.
ElPaso:
Do you think the dynamic memory allocation is the most probable cause then?
I'm not sure. It certainly is an area of suspicion but the particular usage of it in this sketch looks like it could possibly be safe since it allocates the memory, does one thing with it, then deallocates. However, I'm not sure what could happen between the allocation and deallocation.