Optical HRM still totally wrong with 21.19 - In fact, worse than ever

Three intervals yesterday:

It's weird that the HR is correct at the start of the second interval but then immediately drops too again for the entire duration of the interval. 7 minutes in this case.

I downgraded to 15.19 several weeks ago. All modules were also downgraded (sensor module 24.22). This is not recommended by Garmin, but it solved all OHR problems. I have done a lot of running and cycling activities and the heart rate was also accurate at the beginning of an activity without jumps.
Last week I had the bootloop problem and also did a factory reset (had to do this several times).
The firmware is still on 15.19 and all modules have the same version numbers as before.
BUT: Since the factory reset, the OHR problems have returned. I went running two times and cycling two times. At all 4 activities the heart rate was initially too low and I had to turn / remove the watch etc. to solve it. After that the heart rate jumped up.

Garmin-Sierra

This seems to indicate to me that it is not necessarily due to the firmware version. Possibly this is caused by other installed files??? This could also be the reason why it doesn't occur for some people.
This may help with troubleshooting, although I don't know if Garmin even wants to fix this.

Just to contribute what is my typical experience: from my perspective heart rate is heart rate. There should be no activity required. I understand that some smoothing makes sense but it should be able to follow along with a modest delay. Below is today's heart rate 955 with no activity taken shortly after I got off the bike trainer and an Edge 1040 w/Garmin HR Duo strap for the ride. The results are not on the same planet.

Forerunner max hr 111

Edge max hr w/strap 149.

This is a very interesting report. Can you imagine any outer circumstances which might be related to the lower performance of the OHR measurement after a reset of the watch.

I'm still guessing in my case the jumps in the HR reading are related to the ambient temperature and wearing a long sleeve shirt. So the watch does not generate any signal where the HF can be extracted. Watch is mooving a bit and als blood vessels are small at the beginning. As use system is getting to more load all vessels are widening even get swollen so the theres much more useful signal to evaluate.

But assuming Christian888  had similar influences before and after the reset, it could indicate that garmin uses a self learning algorithm to evaluate the HF signal. It might geht "stuck" somewhere since it is not ideal at the end. (Link there seems to be reports for the automatic lactate evaluation.

I have a new example of a run two days ago:

It was about -1°C, I was wearing a long sleeve und got an acceptable result.

So I have to say: I could not find any "real" pattern so far and I do not know if my guesses were right so far.