Shared libraries vs code duplication in microservices

my team is facing the classic shared library vs. code duplication dilemma in our microservices architecture. we have about 20 services, and there's a significant amount of common code: logging patterns, authentication middleware, error handling, health checks, etc. currently, it's mostly copy-pasted, which obviously leads to inconsistencies and painful updates. we've tried a few approaches for sharing. shared npm packages felt like a dependency hell, making deployments tricky when updates weren't backward compatible. git submodules ended up being a maintenance nightmare. some teams just copy-paste and accept the tech debt. none of these feel like a truly elegant solution. what strategies have you found effective for managing common code across a decent number of microservices without introducing tight coupling or significant operational overhead?