The SoC exposed to the user is the BMS's best estimate of usable capacity. It adjusts this on the fly, and under normal circumstances you can be pretty well assured that if it says you have x% available, you have x% available.
If it's inaccurate for some reason, there's generally no way to know until you run the pack down to a very low SoC and the BMS quickly realizes the issue.
It's also possible for things to make these calculations completely useless, but this is generally the result of a third party modifying the pack in a way that makes it impossible for it to make accurate estimates (replacing modules with mismatched ones, snipping cell fuses, etc)... in which case shutting down with even 100+ miles on the dash wouldn't be unexpected. (Yes, this is a PSA to not let anyone butcher your battery pack.)
I would say that you there is ways to detect when the BMS is "off track". Not during a specific drive, but in general we can keep track of this.
When at sleep the true SOC can be measured, but when driving the displayed SOC is a running estimate (calculation) from the known SOC before the drive.
This estimate is initial remaning energy (moninal remaining as per SMT data) minus the consumed energy which gives as the new remaning energy. Divided by the estimated capacity (nominal full pack as per SMT) we get a estimated SOC).
When the BMS is perfectly on track, the estimate and the real SOC is more or less the same.
But if we have a overestimated battery capacity the estimated SOC when parking is too high, and after a short period of time (like 15-20 minutes after parking), we can re-check the displayed SOC and it will have decreased as the overestimation causes a underestimation of the used percentage of the SOC.
This is the other way around with a underestimation. The SOC will then instead increase after the drive.
For charging, the opposite to above happens, charging is estimated by BMS "To charge complete" so a underestimation results in a overshoot in charging and vice versa.
Looking at the SOC graph in for example Teslafi show clearly if we have a over- och underestimation.
I did use that principle fot my model 3 when the BMS somehow was very off, showing 75.7 kWh nominal full pack (a sudden drop that afterwards recovered), and found the capacity to be close to 79 kWh. I did a 100-0% drive (using Scan my tesla and teslalogger for all data) and found the capacity to be 79kWh.
For my new Plaid, it showed NFP 95.3kWh when I got it, but the same calculation as per above said 98.0kWh. After 3-4 weeks the NFP entered 98kWh and has stayed there.