Microservices should be lean and agile. And that's usually true. However, microservices-based architecture usually requires additional overhead. The first layer of overhead is related to the additional interfaces used for microservices communication. RPC libraries and API providers and consumers have to be multiplied not only by the number of microservices but also by the number of their replicas. Then there are auxiliary services, such as databases, message queues, and so on. Those services also include observability facilities that usually consist of both storage facilities and individual collectors that gather data.
The costs that you optimize with better scaling may be outweighed by the costs required to run the entire fleet of services that don't bring immediate business value. What's more, it may be hard for you to justify these costs (both in terms of infrastructure and development overhead) to the stakeholders.
