when does caching become a bottleneck itself?

we've been scaling our read-heavy service, and caching has been a lifesaver. we're using Redis as a distributed cache, and it's definitely helping us handle the load. But I'm starting to wonder, at what point does the caching layer itself become a bottleneck? Are there specific metrics we should be watching, or architectural patterns that signal we need to rethink our caching strategy beyond just adding more cache nodes?