Some Java services don't fail because of traffic. They fail because background jobs don't know when to stop.

A background job should never be able to take down your production system.

Ours did. Every week.

At SAP, we ran a multi-tenant service processing thousands of tenants. Every morning, a scheduled job kicked off - telemetry, cleanup, routine work. Nothing unusual.

Until it overlapped with real traffic.

Then, within minutes:

Heap jumped from 60% → 94%

GC went into panic mode

Latency spiked 5×

OOMKilled

No memory leak. No bad deploy. Just a background job that didn't know when to stop.

If This Feels Familiar, You're Already at Risk

If you've ever:

Used @Scheduled and assumed it's safe

Configured a thread pool and felt "this should handle it"

Relied on auto-scaling to absorb spikes

You're sitting on the same failure mode we were.

The JVM didn't fail.…