The way I understand it, and I could be totally wrong, that’s somewhere between dead and impossible within the span of 2 hours.
You get the occasional outliers from a bad reading but in general that's how I've experienced, and I've had nurses and doctors say that they are "good enough".
And yeah in my experience, anything less than 90-93% in a hospital and they will have you on high flow oxygen pretty quick. So 85% as a single bad reading is fine, but a sustained 85% on a watch means you probably want to head to a medical facility.
There simply haven't been enough (or sometimes, any) tests comparing what are normal/safe readings for continuous monitoring in normal life settings of many of these parameters, especially for healthy people, and doubly especially for these types of non-invasive simplistic sensors.
Basically, we don't know how much should a healthy person's pulse/SpO2/... as measured with a smart watch sensor vary during the day. We also don't know which values, if any, should be considered emergencies, or which values should scare you into a programmed visit.
Note that over-monitoring is often just as harmful as under-monitoring in medicine, especially give that the vast majority of the population is healthy (so any false positive is likely to affect many more people than a false negative does, even if in lighter ways).
See the video for the Apple Watch Series 6 [0], and Series 7 [1].
There's also tests for the Series 8 [2], although it doesn't include data collected in a low oxygen environment.
[0] https://youtube.com/watch?v=8HIcwMhEny0
If I open the app and stay still for the full 15 seconds the reading seems much more accurate.
I can confirm with the other poster, I've found them to only be ~1 or 2 different from other oximeters.
I'd always assume bad read first if you have an outlier.
The main use cases are for detecting sleep apnea and high altitude hypoxia. If you aren't subject to those issues then you can just disable the pulse oximeter to save battery life.