I'm working on a biggish Go application (a rebuild of a poorly written 160K LOC app) and I've been trying to make it "perfect" because it has to work for the next decade, and (in theory) I'm laying the groundwork for a slowly growing team of developers and my successors.

But in practice I'm stuck on the one side in analysis paralysis, and on the other I feel like what is pushed as best practices is slowing me down and adding a lot of boring, fragile boilerplate (use case: adding a field means changing structs and conversion functions in 3-4 packages).

I'm seriously thinking of just flattening everything again and work from my database representation, because right now I'm not moving very fast and adding / changing fields feels daunting. And it shouldn't, because that should be the easiest part, and it's something I've done for so long already.