I suspect SQL Server 2016 runs better than 2005.
It's always an "it depends" discussion. Some things run for years, and there's no reason to change. Some things do need to change because they deteriorate. Learning to decide when change is needed and when it's not is a useful skill.
I think far too many devs want to change because they want to change, or they're bored, or they see something shiny. Or maybe they believe marketing. I think Jeff sometimes avoids change just because, though I do see him query people about why they make a change and the benefits for some things (PoSh), which means we can have a rational discussion.