1. Depending on how you are recording, using midi, one device will have to be the master controlling tempo, start stop functions, etc. The other device will have to set as the "slave:" device. I'm still way back on Sonar 8, but I can't imagine this has changed.

2 ( and 3 ). Uncle Dave and Rikkisbears gave excellent answers. There is also another possible issue. If you record into Sonar with your SX900 rather than using the internal sequencer, when you go to play the resulting midi file back from Sonar, if the SX900 is set to receive midi on the sequencer ( often default ) or other channels instead of receiving midi and having it received by the style and/or proprietary voice / performance channels, you will almost certainly get the wrong sounds. This is often setup this way by default so you can play Standard midi files using standard midi or sometimes XG voices rather than the Yamaha proprietary voices. Otherwise, unless you know the proper cc and midi instrument numbers and have them entered into your sequence, using cc changes and the proper Ins. device, chaos often ensues.

4. Depends on what you have setup in transmit / receive. At least it did on Sonar 8 and my older Yamaha's ( Tyros 2 for example )

5. You'll have to compare the midi data and program numbers from model to model via your .ins files and / or the manuals. Back when XG was the standard, all the Yamaha boards, IF they had XG, were the same in that area, and the "proprietary voices" weren't always exact, but many, like going from the PSR740 to the 2000, were close to beingthe same. With all the extra data from the mega and SA1, 2 etc voices it's possible that changes it a bit or a lot. In my own case, I'd have to compare the T2 to the Genos and see how different they are. I have both but I rarely use external midi devices to record into now.