Answer
This is very commonly caused by the built-in PSO (Planning & Scheduling Optimization / field service scheduling) integration, which by default has a scheduled database task actively running (something like 'Send PSO Integration dataset updates') even in environments that aren't actually using that integration — every run that has nothing meaningful to send, or that fails to reach an unconfigured PSO endpoint, shows up as another failed Application Message. If your environment genuinely doesn't use the PSO integration, disable it directly in its dedicated configuration/setup window (rather than trying to address the failed messages individually) — this stops the scheduled task from continuing to generate new failed messages going forward.