TL;DR: How can I know if the SpO2 reading is in fact taken in optimal conditions ?
As much as I understand your frustration, Apple mentions the limits in the small print at the bottom of https://www.apple.com/watch/:
Blood Oxygen app measurements are not intended for medical use, including self-diagnosis or consultation with a doctor, and are only designed for general fitness and wellness purposes.
I have had similar conversations with people about other aspects of Apple Watch - heart rate not matching expectations, people comparing their activity with another and feeling the calories or heart rate or distance travelled aren't the same, so hopefully some of that perspective can help you and others.
I find the largest value for my health data is longitudinal tracking of my stats with my watch and my band. Yes - an instant value is nice when perfect, but over time, those measurements are easier to trust as you see how they vary.
Each of us has different mechanics, melanin pigmentation, and bone structure. The watch is a sensor and it surely aims to get a good result, but your results will naturally be different from mine. What doesn't change is your results day after day and week after week with the same sensor will give you good information how your body is changing.
Specifically, pulse rate and SpO2 are visual sensors, so band tightness, interference due to body hair, transdermal fat content, skin pigmentation, current moisture levels, and your specific capillary structure and location of anterior arteries and veins on your watch bearing wrist should affect the readings. I'm sure Apple tried to control for this, but the variation in humans is going to make some people have larger error than others. Hopefully you are able to get decent results once you control for the elements you have control over.
This error and confirmation of accuracy happens all the time in physiological labs for any medical measurement. The instruments are almost always very precise (measuring temperature to a tenth of a degree) but not always accurate - three devices all read 37.2° (or 99.0° in a few countries).
Same with estimating your stride length - if you are training against yourself - it doesn't matter if the reality is your steps are 1.4m and it's measures 1.2m if you're just looking at how far you go compared to yesterday and if you went further or faster than you did in the workout yesterday. Basically, don't overlook the value in longitudinal data even if it's not absolutely perfect.
My hunch is nailing the calibration is what's expensive and pricey (like building several redundant sensors). Many instruments that are cleared for medical use have formal calibration procedures, which is one reason the Apple Watch isn't cleared for that level of diagnosis. It lacks the ability to perform a calibration step if you determine it's not precise and accurate enough for you.
Also, this assumes the watch is perfect and the sensors are perfect. If over time you can't get a good pulse read / SpO2 read - maybe Apple would exchange it, but I'd hate for you to go through a swap if the device is working as designed and it's just not what you need where your body and Apple's design meet.