It is parallel, but not all jobs are run on the same hardware/software and you can't re-run jobs which error out. In fact, some jobs don't error out but should

It is parallel, but not all jobs are run on the same hardware/software and you can't re-run jobs which error out. In fact, some jobs don't error out but should be invalid and if discovered as such might cast doubt on the authenticity of other jobs, etc. etc. So it's really not as trivial as you make it out to be, and neither is anything else which involves coordinating 100M+ people doing a particular task.