Yeah I think it's trauma due to C/C++'s awful warning system, where you need a gazillion warnings for all the flaws in the language but because there are a gazillion of them and some are quite noisy and false positives prone, it's extremely common to ignore them. Even worse, even the deadly no-brainer ones (e.g. not returning something from a function that says it will) tend to be off by default, which means it is common to release code that triggers some warnings.

Finally C/C++ doesn't have a good packaging story so you'll pretty much always see warnings from third party code in your compilations, leading you to ignore warnings even more.

Based on that, it's very easy to see why the Go people said "no warnings!". An unused variable should definitely be at least a warning so they have no choice but to make it an error.

I think Rust has proven that it was the wrong decision though. When you have proper packaging support (as Go does), it's trivial to suppress warnings in third party code, and so people don't ignore warnings. Also it's a modern language so you don't need to warn for the mistakes the language made (like case fall through, octal literals) because hopefully you didn't make any (or at least as many).