Scalable agents configuration for scraping profiles

[devmentality]

Hi!
I'm trying to integrate pyroscope to my company's kubernetes infrastructure. We would like to deploy grafana agent to scrape profiles using one of the described approaches: https://grafana.com/docs/agent/latest/flow/setup/deploy-agent/
Our challenge is that we have hundreds of independent teams. And they want to be able to configure set of profiles and scraping interval for their particular service, in runtime if possible.
Could you, please, give us advice for the most appropriate deployment strategy for our case?

1. The recommended way is to deploy "As a centralized collection service". Am I right, that then all the agents will share the same config and share work using clustering? Should we make one big config file for all the services with own kubernetes.discovery & pyroscope.scrape sections (https://grafana.com/docs/agent/latest/flow/reference/components/pyroscope.scrape/#example)?
2. Or for our case it's better to run something like "agent per service" and configure agents separately? It doesn't seem like a big problem, but may be more costly than centralized approach.