The other day something curious happened that made me question how I use bang methods (and their non-bang cousins). You see, I wanted to merge some hashes and the performance difference between the bang and non-bang version of merge was astounding, so I decided to dig further…