Each action that you perform on the watch, be it switching through widgets, logging an activity using GPS or listening to Spotify will draw a specific current from the battery which in turn drops the voltage in that moment, only to have the voltage rise again once the watch is idling (i.e. doing nothing more than displaying the watch-face and running the HRM LEDs).
What the watch ends up with is a plot of various voltage levels over the course of a given time period which it will compare against a model to approximate the percentage of charge remaining.
*
* *
* *
^
v *
* *
time -->
"Post Workout Battery Drain" - a term which I've seen used a number of times in this forum, is what people are using to explain the value smoothing which happens at the tail end of an activity when the voltage jumps up following the various circuits being turned off and apps being stopped. Garmin appears to have altered this model in the latest firmware release to more accurately reflect the true state of the battery charge during activities and also following them.
You might notice that your battery burn rate appears higher during an activity than it used to be, but it's not actually the case. What you're seeing is a more accurate representation of the batteries true charge.
And as a final thought for those experiencing poor battery performance - take a look at the apps you've installed on the watch, specifically third party watch faces. I've noticed that certain watch faces that present a large number of metrics can incur discharge rates of up to 1% an hour. This is most likely due to the watch face (it's actually an app) reaching out continually to internet based api endpoints to update weather metrics etc. Something to consider before lashing out at Garmin.