From One Horizontal Scaling Controller to Many: Evolving Uber's Compute Platform

Introduced the ServiceScale CRD and Service Scale Controller (SSC) to manage scaling intents from multiple orchestrators (e.g., normal operations via UDC and failover orchestrator). This design choice avoids overloading the existing Uber Deployment Controller (UDC) and creates a scalable model for future scaling use cases. Addressed production challenges related to Kubernetes informer cache staleness with a read-your-own-write consistency guardrail and developed a fleet-wide observability and automated healer for metadata-spec drift in multi-writer scenarios.