Currently on 2017.07 , but this has been happening for quite a long time now through various versions and I'm not sure where it started, but I am sure that it didn't always do this. If a softsynth midi clip starts at say 08:01:001, the frozen audio clip might be positioned at 07:04:959 or 08:01:005. It's a minor shift, but irritating. I've been altering the clip start time manually all this time, but I'm really fed up of doing it now. 
 
Is it just me ? 
 

I don't know if those variances are illustrative or actual measured examples, but I would be surprised if the audio appeared before the MIDI note triggering it. I would not be surprised if it appeared slightly later - the same as if you recorded the synth's audio track while playing back its MIDI track, due to the instrument's processing delay between receiving the note-on data and generating the actual note. But that's just a theory, I don't know for sure.

Anderton
I don't know if those variances are illustrative or actual measured examples, but I would be surprised if the audio appeared before the MIDI note triggering it. I would not be surprised if it appeared slightly later - the same as if you recorded the synth's audio track while playing back its MIDI track, due to the instrument's processing delay between receiving the note-on data and generating the actual note. But that's just a theory, I don't know for sure.

 
The "before" example was an actual example. One instrument track only , synth = TTS-1 Piano preset, one note. Clip durations also change, which I understand due to release settings, but it was the fact that sometimes clips are early that made it seem to me like something is wrong.
 
 
 
 

Again, just going with theory here because I'm not having this problem...maybe the ASIO interface is reporting the wrong latency? Some interfaces are very accurate in their reporting (IIRC Line 6 is really good about this) while others...not so much. Perhaps SONAR is compensating for latency based on what it thinks the compensation should be?

I think this might be a good question.  I added TTS-1 as a simple instrument track and put one note at 8:01.

(full sized link https://snag.gy/NjMF9a.jpg )
 
Then I froze it.  Notice the start time of the clip, 7:04:959

 
(full sized link https://snag.gy/Hli8Z6.jpg )

I never use TTS-1 and in trying to replicate Matt's experiment it crashed SONAR twice. So I repeated the exercise using the free Synth 1 synth solo'd in another project. I placed the note at 8.01, then froze the synth. The inspector showed 7.04 exactly as it did for Matt. However, the resulting audio clip starts at exactly 8.01 here. When I unfreeze the synth the inspector shows the clip correctly at 8.01.
Seems it's an error in how the inspector reports frozen clips.

Does the frozen audio sound correct? The clip itself may start early based on a number of factors, but the events should be the same as the triggered midi/audio.

To make track importing (into other DAWs) i often extend the first clip all the way to zero. Some synths seem to render beginning exactly on the first event while others render begins a bit of time before the first event?

Keni - I assume you are asking the OP. But just in case, in my case everything sounds fine with the resultant audio. And the audio appears where I expect it to be. The only peculiarity here is the reported clip location in the inspector which I have never observed before. It seems to me that Paulo's system is misreporting ASIO latency as Craig suggested. Not sure.

Well that's what, one tick of time? I'd try zooming in and see if there's any audio there. The clip might have a 1-tick lead in before signal actually occurs.

Yes Brando...
 
I was responding to the OP...
 
He specifies where the clip is early, but not the actual event and as he notes, the ends vary based on the sound's release regardless of MIDI value...
 
I frequently see differences in frozen audio clip start times that vary synth to synth...