Issue Summary

We identified a critical crash pattern in Firebase Crashlytics reports, where a small subset of users experienced an unusually high repeated crash. All crashes occur during the application creation. We found the same pattern in multiple applications in our company.

Firebase Crashlytic Report Pattern

Those crashes happen because of device-specific issues (incompatible Google Play Service or corruption). The reason why the app crashes is understandable, but the main problem we are looking into is why the app repeatedly crashes & relaunched.

Investigation

Observations

Extremely high crash frequency per affected user but small number of affected users
All crashes consistently occurred during app launch
Crashes showed a pattern of immediate app relaunch

Reproduction Steps

To investigate the issue, we:

Create a demo project and include Firebase Crashlytics & Firebase Analytics
Simulated a crash during app launch
Observed the following pattern:

App crash during launch
Automatic relaunch
Immediate subsequent crash
Pattern repeated utils the Repeated-crash-dialog appear. Most of the time, it's three crashes in a row, but sometimes, our demo project shows 20+ crashes in a loop if the dialog is not shown.

Root Cause

The investigation revealed that: While the app crashes during app launch, it is scheduled to re-launch the JobInfoSchedulerService after a period, and Firebase used it.

ActivityManager Scheduling restart of crashed service com.example.background/com.google.android.datatransport.runtime.scheduling.jobscheduling.JobInfoSchedulerService in 1000ms for connection

I understand that JobInfoSchedulerService is a JobService. If a JobService hasn't completed its work (returns false from onStopJob()), the system will automatically reschedule the job for later, and it causes the repeated relaunched.

Since crashes happen during Application creation, any scheduled jobs, broadcast receivers, or push notifications might trigger this crash loop.

Our questions are:

Is our analysis of the JobService behavior correct, particularly regarding its auto-rescheduling mechanism?
Has Google acknowledged any similar issues report?
Would you suggest a solution to prevent such crash loops, especially during app initialization?

Resource

demo project with a fake crash during the app launch.
bug-reports.zip
crash-logcat.txt
The emulator was used for testing emulator_spec.txt

Issue 391763850

Links (6)

Issue metadata