> The user doesn't care if technically there is some trick that is being used to "cheat."
You framed the "benchmark" around performance, not perceived performance. Those are two very different things. We're on HN, so don't be surprised when people correct you or nitpick. Either accept the clarification (if it is correct), clarify what you originally intended to say (without being overly defensive), or move on.
Besides, if we're talking about average user, they're not force closing apps and the phone should be caching the most frequently used apps. So even then the test won't mimic real world experience even if it's about perception. Maybe initially, but not after sustained usage. People usually normalize to whatever they're using
The device should be absolutely caching the most recently used apps, yes. Force closing is unnecessary, but I manage my apps on my phone not because I want to close background apps, but because I want to keep certain apps running (I don't care about steam or my email or messenger, etc, they open fast; the browser is slow and I want that to stay loaded).
Unfortunately, I can dismiss/force-close all apps on my phone, open Firefox, switch to the home screen--and when I switch back it has to start up again fresh about 25% of the time.
The only reliable way I've found to use termux (as an ssh client, not anything high-resource) and Firefox at the same time is to use multi window, dragging termux out of the way of the keyboard every time I switch, as Firefox will regularly need to be reloaded otherwise; if I'm on certain heavy or lazy-loading websites or using a link with a hash in it, it'll then also lose my scroll position. This is particularly frustrating when I ssh into minimized containers with no command completion or man pages.
Besides, if we're talking about average user, they're not force closing apps and the phone should be caching the most frequently used apps. So even then the test won't mimic real world experience even if it's about perception. Maybe initially, but not after sustained usage. People usually normalize to whatever they're using