Longevity in any mental-heavy profession depends on the genuine interest and exercises, even hobby.

Exposure to different experiences with other libraries and frameworks are only good if there is anything fundamentally new in them. For example, concurrency vs single thread execution, different paradigm (functional vs imperative), etc. Thus, experience in Ruby + Python + JavaScript matters less than lets say C + Erlang + Prolog.

It's important to learn general patterns, and the rest of the memory can be used for the specific syntax / framework at hand. With AI-powered approaches syntax becomes even lesser concern so ability to think is actually highlighted. And in that sense we see a lot of examples of thinkers (scholars, philosophers, even religious leaders) who can go quite old without loosing that ability.

Seeing the bigger picture is something many young guys need to learn. Older guys already learned.

As for software engineering becoming more technical. I doubt it. It will become less technical. Most of us do not program in a very technical machine code or even assembly languages now a days. Yes, there are some bodies of knowledge, which grow, but AI will help navigate them, so at the end IT will grow a good metalevel, hiding even higher level technicalities. Thus the level will become even more humane.

The biggest problem for humans is to manage complexity. Even though complexity will not go down, the ways to manage it will improve drastically in the coming years.