Why Are My Two AWS Fargate Tasks Performing So Differently?

Totally agree with what was mentioned earlier! If you really want control over the instance types, check out AWS ECS Managed Instances. It's a great alternative because it allows you more control compared to Fargate.

One reason for the performance difference could be that Fargate doesn't guarantee which CPU chip your tasks will use, leading to possible performance discrepancies due to different hardware. If you're dealing with tasks in different data centers, that might also cause variation in performance. You might want to switch to the ECS EC2 launch type for better control.

You're on the right track about round-robin routing possibly not working as expected in Fargate. Switching to a Least Connections setup might help balance the load better across your tasks, especially if some are underperforming.

Fargate isn't designed for consistent performance. In my experience, if you need fast and reliable processing times, going back to EC2 might be the better choice since you can select specific instance types there.

The fact that one task takes so much longer suggests there's some internal or external factor at play affecting performance. Have you instrumented your app to monitor things like dependency times? Tools such as Datadog or New Relic can provide insights that might pinpoint the delay.