In a similar position, I don't write much code anymore, but I review a lot.

I don't do the 'you should have made this const' or 'this variable is badly named' type of review... But I do a higher level check that nobody is overcomplicating simple things and building 37 new microservices when they ought to just be using one that already exists.

Overcomplicating simple things is a really common way to justify growing your team, getting a promotion, etc.

Suddenly the 'internal printer management team' needs 8 developers, dedicated network bandwidth, and a project roadmap multiple years out with shiny webUI's, when all we really needed was a couple of dumb printers and 0.1 people to manage them.

GitLab team member here (the one who was shadowed). Most of my merge requests were for the engineering portions of the company handbook.