I have "looked it up" and I have designed medical PPG devices professionally also. They absolutely do not "estimate" your heart rate any more than holding your finger to your wrist and counting your heart beats in a minute. That is also "estimation" just like literally all physical measurements that are measured for less than t=infinity are.
Of course, there are manufacturers and models that do it badly, especially during weightlifting or heavy flexion/high wrist impact activities.
1 lead ECG can also very very reliably measure your heart rate even during and that is what chest straps like the Polar H10 do in a variety of exercises.
Here is a good overview of fitness watch PPG testing in a variety of scenarios and how it correlates to 1 lead ECG. This guy does tons and tons of extended tests. I just chose a recent one that has recent devices. Even sleep cycle recognition in modern smart watches is getting better and better correlation to EEG, which this guy also tests sometimes.
Stress, recovery times, VO2max, etc.. All of the "1 degree removed" statistics are generally a loose educated guess and generally crap.
Heart rate, HRV, skin temperature, speed with GPS + IMUs, and to a certain extent relative activity intensity can be directly measured and are not a guess if the device is designed and tested well (which admittedly, only like 10% of devices are)
Blood pressure measurements using the principle of delay between electrical signals and pressure wave elasticity is also being FDA certified right now as accurate to a certain percentage of medical pressure cuffs.