Sounds really strange, but the info you provide imo is to limited to expect any helpful answer.
If FHEM is crashing "reliably", you should analyse the FHEM log (and perhaps try to get an extended log) or start FHEM in a special debugging mode and try to reproduce the problem.
(May not solve the underlying problem, but) why are you not just changing the DEF of your original GW to point to the new one?
Additional questions you may answer:
What type of GW did you use and are you now using?
Controller Hard- & Software (Pi or decent Linux machine, Windows server, VM...???)
Does FHEM also crash, if you try to autocreate the problematic Node not as #11, but as the first?
How often died you save your config in the process of including nodes? Did you monitor memory usage?
If crashing FHEM is related to a "special" node: What is different in this sketch?
...