Since 2013, #curl has had more lines of code/data in git for doing tests than we have production code.

We are now at 1200 lines of test infrastructure per 1000 lines of production code. And we still could do a lot more and better testing really.