The pprof tool has many different methodologies for incorporating profiling into your code. The Go language creators wanted to make sure that it was simple and effective in implementing the profiling necessary to write performant programs. We can implement profiling in many stages of Go software development—namely, engineering, the creation of new functions, testing, and production.
It is important to remember that profiling does add a small performance penalty, as more metrics are being collected on a continuous basis in your running binaries. Many companies (Google included) feel that this trade-off is acceptable. Adding an additional 5% overhead for CPU and memory profiling is worth the cost in order to consistently write performant code.