I just figured I would come chime in here on this thread.
For those who doubt and want to defend Sonar as a good midi sequencer...
Sonar is NOT following midi sync correctly.
How can I say this? Because I wrote my own midi sync source using nAudio and c#. My source plays Yamaha style files. The START (0xFA), STOP (0xFC) and CONTINUE (0xFB) events function correctly, but the clock event (0xF8) is *NOT* implemented correctly, and, as everyone on this thread has already stated repeatedly, Sonar is basically ignoring the clock events and using it's own internal clock for the timing of the midi events themselves (regarding beats and measures). How do we know this? A simple look at the Piano Roll reveals all (as the timing is very clearly not synced). The best test is to make no manual attempt to synchronize the tempos between the source and the Sonar project. In fact, make them wildly different, just to illustrate the point. The tempo in Sonar should have no impact on anything. In fact, in slave mode, it should be greyed out.
But then the developers stand up and say "Not so! The midi clock events are NOT being ignored!". =) Well... this is true. So long as I leave my Sonar sequence in slave mode for the rest of all time, all "appears" well with the world. I struggle to understand why I would want to do this, though, since I can't quantize anything in Sonar this way, but Sonar does drive playback and recording from the external clock events with each recording pass. But... with just a little extra attention to detail, the implementation could have been written to line up the notes to beats. Why would we want to do this? So that we could eventually cut out the external clock and run strictly with our sequence as it exists in Sonar. It is truly sad to see midi being treated as some kind of inconvenient afterthought by Cakewalk.
So then... in this experiment of mine, what is my "control"? To what can I compare Sonar to show that they are doing it wrong and someone else is doing it right?
Well... believe it or not, I'm the last kid on the block that is running Windows 7 32 BIT expressly for the purpose of continuing to run Cakewalk 3.0. And miracle of miracles, when I conduct my exact same test with Cakewalk 3.0, it works, and all of my drum beat notes are nicely lined up on the beats. I turn off the midi sync and run the sequence and I am golden for coming in and laying in some additional tracks in Cakewalk itself (and being able to quantize later).
Now... I realize that Cakewalk 3.0 may represent the last time that cool kids actually worked at Cakewalk... and judging from the complete lack of follow up on that EXCELLENT midi equivalent of notepad, we might assume that the cool kids left the company and didn't leave any backups of their source code. However... if this is not the case, it seems to me that the current developers could go dig up the old code and find out what the old timers had done right.
Just from reading the comments on this thread, it shouldn't be confusing what people want it to actually do, and there is no really good defense for why it doesn't.