Microservices scaling: independent vs coordinated deployment strategies

in our microservices setup, we preach independent scaling and deployment. but in practice, i've noticed that scaling up one service often requires scaling up others in lockstep. for example, if we have a traffic spike on our api gateway, we almost always need to scale up our authentication service and user profile service at the same time to handle the increased load. this feels like a coordinated scaling anti-pattern, but it's often the practical reality. are we doing something wrong in our service decomposition or resource allocation, or is this a common experience for others running large microservice architectures? how do you genuinely achieve independent scaling for closely coupled services without manual coordination during peak events?