We don't work with take lanes.
But this brings me to an idea:
Now I have unchecked "Snap to nearest Audio zero Crossing". Maybe the zero crossing generates the problem. It may be that Sonar cuts the clip at the very left side at a zero point - instead of deleting the whole portion. Although then the microclips would appear more often.
I will check this.
 

For a test, you could also keep "Snap to nearest Audio zero Crossing" checked, but set "Selection after single split" to "None".
You have to select the left side of the clip you want to delete manually and that's more work and maybe not feasible for you, but it should ensure that the clip portion is selected entirely.
 

Just happend again. 
I can reproduce it. 5 times tested.
No matter if "zero crossing" is checked or not.
No matter if "Selection after single split" is set to "None" or "Left portion". Even if I delete the selected clip manually.
Now I can reliable reproduce that the deletion of a selected clip leaves a microclip at his left side.
   
In this case 2 recorded tapes consisting of 4 clips (each tape has two sides).
With a full length of 4 hours and 22 minutes.
 
 

I must correct myself.
The microclip is NOT the left side of the clip.
The microclip remains to the RIGHT side of the previous clip.
That was not to be seen immediately.
 
That is the evidence: (zipped 855 MB)
I have uploaded the project:

https://office.1und1.de/guest?path=K18936%20-%20info%40hambrecht.de&token=4849536D40C10A33&mandant=02&product=onlineoffice&locale=de&viewType=0&guestToken=cRgIUshXTuOTh2dBK1-7nw&loginName=info%40hambrecht.de

 
I have shortened the project to 1 audioclip:
- Just split at the end of the music at measure 1977.
- Delete the right portion.
- You will see a microclip at measure 1986
 
Fully reproducible - no coincidence.

I'm downloading...

I just looked at your project:
There is a second Lane (expand Take Lanes to see it) which contains the remaining portion of the clip (although in the same Lane).
 
You don't have to split the clip, the remaining microclip is already there.
 

I think this is a remaining of a not "Flattened Comp" track.
 
 
Edit: In this case it's faster to just delete the microclip than to use "Flatten Comp".
 

--- deleteted ---

Thank you a lot.
I have found the reason. I was completely wrong.
I record the first side of a tape as first clip. Then I stop the recording. 
Then I turn the tape and record the second side as second clip.
The problem is that the two recordings are "now" a little little bit overlapping. That was not yet a few months ago.
So this problem comes probably with Comping and the new take management.
 
It should not happen that after "stop" and the next "record" the clips are overlapped. But now I know that I must leave 1 measure empty to begin the next record.
 
I have been annoyed for months.
But without you and this forum I would never haved solved this problem.
Thank you.
 
 
 

I opened your test project again and actually you don't have to expand Take Lanes.
If you zoom way in you'll see the (muted) microclip right at the end on the main track.
 
So, the second empty Take Lane is probably not important, but a previously split (maybe inadvertently) seems to be done and since the remaining clip is also muted (which is not the case of usual splitting), it looks like "Comping" has been used.
 
But it would take a couple of inadvertently pressed keyboard shortcuts to get this result, so it doesn't look like an oversight or user error.