I'm curious how many people are testing their application performance and usability over these kinds of connections (low bandwidth, high latency or lots of packet drops), and the extent to which the results of that testing lead to changes in the applications.

And if you are doing this kind of testing, how do you do it? Simulating a poor network connection with something like the sch_netem module (https://openwrt.org/docs/guide-user/network/traffic-shaping/sch_netem), the built-in browser tools, actual low quality connections, or something else? Is this a regular part of your testing process, or a more ad-hoc thing?