Orchestrating microservices in geo-distributed Cloud-to-Edge environments introduces challenges, particularly with maintaining response time SLOs due to network instability. Kubernetes, the standard for container orchestration, faces limitations in scheduling and descheduling when applied to such complex, distributed applications. Its default policies do not account for response time SLOs or the dynamic state of applications and infrastructure. This study proposes an enhancement to Kubernetes with a network SLO-aware scheduling and descheduling strategy, enabling adaptive placement based on the observed runtime application response times. Our approach is benchmarked against default Kubernetes scheduling policies to demonstrate its effectiveness.