We are using GA4 in our Android app and we recently switched from automatic to manual screenview tracking. So we switched off google_analytics_automatic_screen_reporting_enabled in the AndroidManifest and started logging GA4 FirebaseAnalytics.Event.SCREEN_VIEW events.
By the same logic how we have been doing (and are still doing) it in Universal Analytics without a problem.
The issue is that with that release the engagement times in our GA4 reports have drastically gone up - compared to the previously automatically captured times for screen classes, but also the overall average engagement time. Instead of something like 2mins for a screen it now shows an average of 15 or 30 minutes or even 1 hour.
The numbers didn't change for the iOS app where we introduced the same changes, so I'm a bit puzzled what might cause this. The manual screenview tracking is my best guess, but it could potentially be something else as well.
(Brainstorming about potential causes in regards to the screenview tracking in GA4, we also looked at the fact that we sometimes call logEvent in some RxJava stream on a non-main thread. Wondering if this by itself or a slight delay caused by it could somehow mess with the measuring of the engagement time under the hood - possibly something lifecycle-bound?!)
(Another thought has been that the GA4 Android SDK doesn't properly handle when the app is put in the background without automatic screenview tracking. Maybe it's measuring screens until the app is eventually destroyed by the system and altering the averages that way.)