It is look like they removes a gap between IR commands in a macro (and, conjointly, the gap between RF preamble and RF demodulated code). I can't test this now due I have no 950, but if it is true, it means the restoration of old Pronto timing strategy, lets rejoice!!!

First, it allowed (again!) the Long-Button-Press as
[A] Code
[A] Code
[A] Code ....

Second, the trouble with devices that sensitive for this gap is solved easily:
[A] Any-previous-code
[D] 0.2s
[A] Sony code
It forces to inspect the macroses, but this sacrifice is defensible die the first possibility.

Third, the problem with x10 through RF is solved by inserting a gap to a command - instead of command
0000 006B 0000 000C 00A5 0096 002E 0105 008E 0096 002E 0105 002E 0105 008E 0096 008E 0096 002E 0105 008E 0096 008E 0096 002E 0105 01CA 0DF5
it is need to use
0000 006B 0001 000C 002e 0DF5 00A5 0096 002E 0105 008E 0096 002E 0105 002E 0105 008E 0096 008E 0096 002E 0105 008E 0096 008E 0096 002E 0105 01CA 0DF5
If it still not worked well, it wills a reason to increase the gap between command repeating parts: 0DF5 means 0.09s in this command, and it is too short for x10, IMHO. Lets use 3000 and more instead of 0DF5!

If it is really happen, the not so many things still need for me to be sure that I can control any device from NG:
- The toggle capacities for 7000 codes
- Possibility to inhibit IR code optimization in importing/learning (as possibility to overcome the misses of this optimization)