It's a cost benefit analysis, if the failure costs you more than verifying it then you will verify it. Since cost of verification is currently high, if you are not losing millions due to failure it doesn't make sense to verify your code.
Look at fuzzing. No one did fuzzing a decade ago. Since fuzzing became cheap (i.e. due to cheap compute and somewhat due to cloud computing) everyone does fuzzing on pieces of code which remotely appear to be critical.
Even testing. When cost of running a program was high people didn't write test cases. They hand checked programs for bugs.
