@M@
Our turntable is a broadcast turntable EMT948. "Supposedly" the best of the world. Has been used in all european broadcasting stations. It runs more accurately as a switsch watch. Second hand (with system TSD15) 6.500$
 

Nice!

I can't imagine an overloaded HDD (wrong cache flushing) would create seamless and error free .wav files but that should be easy to check.

OK here is the first TEST:
SONAR looks into the future:
The recording is ready before the end is played.

 
The ghost is SONAR. At the end of the recording, a ghost from the Cakewalk team comes immediately and cuts a hole and joins the ends. Probably everyone laughs at this "joke".
(I have turned off direct monitoring. The output of the turntable is synchron with the RME 800)
 
My next test is updating from 2017.02 to the current version.

Ahhhhhh firewire...... although we love it when needed, continuous placing into new set ups is risky so why continue...
This is copied from multiple current trade rags,
Elegy for a dead technology

"Today FireWire is fading into memory. Thunderbolt took its place at the high end of the market. And at the volume end, USB 2.0 has given way to the much faster USB 3.0, which is now being replaced by USB-C.....etc etc.
And Dell budget laptops for audio commercial set ups, why use Dell ?

And why put multiple set ups out in the field when there were no running systems being tested on the bench before putting in production?
Make some hardware changed with known good systems, solving the problem, before hundreds of hours are spent in the weeds.

If you are recording with RME DigiCheck and you are messing with record enable buttons in sonar, you can lose what was in the buffer. You risk the data in the buffer, so bigger buffers means more obvious skips.

This is basically RME driver issue. Don't fiddle with the driver while you are recording and you will not have the issue.

Both Sonar and DigiCheck can be tricked this way.

When you are multi tracking, all tracks are impacted identically.

@Joe_A
You are right. I end the tests.
We had for years a working computer that records withe the RME 800 with no problems. We only wanted to save energy, so we took a laptop. It is difficult to find a laptop with Firewire, so we took the DELL.
Therefore the DELL has to go - and we will continue to use the computer.
Bye bye DELL, you get another job with an USB interface.
 
But the question remains: Why does SONAR not notice that there is a piece missing?

@gswitz
We don't use DigiCheck and we don't fiddle with the drivers. We noticed that "all" recordings from the last weeks are unusable. Because all recordings have holes.
This does not explain why 2 seconds are missing and the record is 2 seconds shorter as the input recording.
Sonar records a stream of 3:17 and has a clip with only 3:15. I would understand, if there were 2 seconds of silence or crackle, when the buffer of the firewire delivers nothing or nonsense.
But a recorded stream of 3:17 must end in a clip with 3:17. Because the RME 800 tells no time code, so the recording of the stream must go on. Even when the firewire delivers nonsense.

IMHO the answer is in one of the circumstances already mentioned above, but which one? Or one very similar.
Targeting that as a different problem and intentionally spending time there is a project for someone.

I think you've done well in resolving the Customer problem getting back with known good systems. And as we all do, learn and move forward....

Thank you all.
This problem can not be solved. Otherwise I - or the Cakewalk team - or DELL - would have to do a lot of work for testing.

This would be appropriate in the RME forum too.

Personally, I think your problem isn't with Sonar. I think something is messing with your driver on that computer while you record. Human or code, idk.

For example, could concurrently watching Netflix cause this intermittently? Maybe.