As noted in post #24, I was not able to reproduce a problem with no notes between the controller and the point at which playback is started.
 
I have not tried adding complexity to the test project as it seems like too much of a wild goose chase with a low probability of success.
 
As I see it, the bottom line is that the searchback-on-render 'issue' needs to be a feature request, and the playback issue needs a solid recipe that can be consistently reproduced, preferably with bundled or freely downloadable synths.
 
I'm not saying I don't believe it can happen; but this thread can't move back to the Problem Reports forum until it can be consistently reproduced.
 
BTW, I noted along the way that there is a problem with Freeze not rendering the effect of a controller back at 1:01:000, but Bounce worked as expected when the controller was included in the selection so that seems to be a separate issue.

azslow3
But @brundlefly:
* have you tested the visualization problem, for which I wrote exact instruction? (post 21)

Yes, I've previously indicated I'm aware of the controller shadow drawing limitation but have not seen any correlation with actual playback behavior. Controller Searchback existed long before controller shadowing was added, so I would not have expected the display issue to have any bearing on the playback function.
 
In any case, I'll try your recipe using CC 28 to control Bass Tone in TTS-1
 
I'm curious whether you've tried that with a VSTi?. TTS-1 being a DXi and having gone so long without updates would not normally be my first choice for testing controller response.
 
EDIT: Did a quick test and verified your recipe. But if I'm understanding correctly this only happens if you deliberately set the beginning boundary of a clip sometime before the first event in the clip. That's a pretty far out scenario that would never happen in my world. And right offhand, it looks more like SONAR is sending Zero Controllers because of the null clip lead-in, rather than that Searchback is failing. If it were just searchback failing, I would expect the control to stay where you put it before starting playback (e.g. centered) until the first event in the clip came up, but it snaps to -12 regardless.
 
EDIT2: Had another look using a MIDI diagnostic plugin, and it appears in this scenario that Searchback is indeed not executed no matter how recent the event, and the value of the first controller later in the clip with the artificially early boundary is sent wherever playback is started within that empty region of the clip - essentially a 'Searchforward' - unless there's a note event some time before it. Worth reporting, but kind of an obscure scenario, and, as you said, probably does not account for Mr. Copper's issue.
 

azslow3
So the only question is who from us is going to submit (2) bug reports?

Good question. It will have to be someone who both cares and has the necessary attention to detail and powers of description. I think that's you in this case. 

Only if you want to change the title to something like "Controller Searchback Fails When Starting Playback in an Empty MIDI Clip Intro". Your issue with there being a limit to how far back Searchback reaches has not yet been replicated.
 
And I think the shadowing issue is totally unrelated to any of the performance issues, and should just be separate thread and CWBRN. I didn't find a relevant thread in this forum, but it's been discussed enough that I would guess a CWBRN has already been submitted.
 
 

azslow3
As I can see in the fixes list, shadowing issue is already address in the next update.
 

Are you talking about this?
 
      PRV MIDI CC info could disappear under certain circumstances
 
I suspect that's a different issue. There are situations in which the controller "tails" themselves can disappear, not just the shadows. But I suppose it's possible they addressed the shadowing issue at the same time.