AWS ECS users face a difficult trade-off when deploying spot instances for small-to-medium container clusters. While spots offer 60-70% cost savings, they terminate frequently, triggering alarm fatigue and brief service interruptions for workloads with fewer than 50 tasks. The core issue stems from ECS's limited placement strategies.

Using a 'spread' strategy across instances prevents widespread outages but forces massive over-provisioning, often negating any cost benefit. Conversely, 'binpack' strategies maximize resource efficiency but concentrate tasks, meaning a single spot termination can impact 30-40% of capacity and trigger customer-facing errors. Unlike Kubernetes, which allows granular control over pods per node, ECS lacks a middle ground for balancing density and resilience.

Engineers are left choosing between operational noise or eroded savings. Common workarounds include using hybrid on-demand base capacity or physically constraining density with very small instance types. Ultimately, the platform's binary options force a compromise that scales poorly for the majority of modern microservice deployments.