Invalid Total Descend

It could be due to decimals and rounding 

First, if the barometer based ascent + descent are within +/- 1 m, the discrepancy in this case is only 0.3 %, which I would call incredibly good! Also, it's difficult to say (because Garmin doesn't tell) what counts as ascent/descent and what is still considered "flat". That could easily cause a 2 m difference during a 10 km run.

(I notice that your activity shows an ascent of 1 m during the descent phase at 32:36. That's not important, but could be either real or just caused by air pressure changes / rounding.)

1. As i mentioned before, for longer run - difference higher.

2. If it due rounding - then rounding should be implemented in same way for End Elevation also.

There is not barometer issue. This is issue in calculation. End Elevation with Highest Elevation obtained in same way. So, we don't care about barometer precision.

And how much is the air pressure changed during your activity?

(but my real question would be, really, you are annoyed by 1 meter high difference?)

"And how much is the air pressure changed during your activity?" Who cares ?

" really, you are annoyed by 1 meter high difference?" did you read last sentence ? I put this example because it is easy to see bug. In other activities - difference higher.

connect.garmin.com/.../11154274760

But garmin shows:

Sounds about right to me. Not exactly right, but about right enough. 

What do you expect to gain by more precise numbers?

Could also just be the principal. Garmin has the numbers and should do the programming correct. But there is a real problem. Garmin programmers aren’t that skillful. About 1% off isn’t a bad score for garmin software. 

Write a function which perform simple arithmetic operation like addition and subtraction - it is not need to be a programmer guru.

"What do you expect to gain by more precise numbers?" Im expect that garmin will fix this bug, which is come at least from 920xt. I think 8 years more than enough to find and fix.

"About 1% off isn’t a bad score for garmin software. " - Yes, if we will compare absolute values versus garmin reported values. But, if in this case we comparing garmin reported values versus garmin reported values. In this case 0% errors allowed

It is like you ran 10km in 1hr with average cadence 160, but got  9550 total steps

I reported a bug that syncing your watch with garmin express, it deletes all routes from your watch. .6 years later, that bug is still there, or it is back.

Since I replaced my old 935 for a 965 I filed a dozen bug reports. Owning a garmin device is living with a *** load of bugs. Other brands have other drawbacks, though

Like I said, garmin programmers aren’t very skilled and garmin isn’t very interested in solving bugs.

And this is not the place to report bugs. You should report it via support.garmin.com. Although it might help to bring it here in the open. But it is definitely not a certainty garmin support notice your complain or do something with it.

My personal opinion, your bug isn’t the most annoying, but I can imagine for you it is.

You my friend have fallen victim of a logical fallacy: The thing is, your watch doesn't know for sure your loop ends at the same elevation as it started. Otherwise I agree they could do your "quick maths" calculation. Even if GPS coordinates of start and end are close, it would not be correct in general to assume same elevation (e.g., somebody running up the stairs of a building, or climbing a rock face).

So, since you watch cannot know this, it has to *measure* the ascent and descent individually, using the barometer. If there are any changes in pressure then you will end up a bit lower or higher (according to the barometer) than your start point, giving you some extra elevation or descent, depending.

Obviously, anything that is measured has some precision/error involved, so what you are showing is normal, logical and also damn close as has been pointed out by others.