Are you attempting to hint towards the use of Cloud Functions/Lamba?
In my experience those solutions and the tooling built around them are still quite immature for our needs.
Also we’re not using a traditional cron, which is often a scary single point of failure, but an App Engine cron, which does not rely on any single instance to do the work.
Finally, traditional autoscaling would not help our situation as it expects a somewhat gradual increase in traffic over time to add the appropriate instances. In our scenario our traffic spikes dramatically, requiring at least 10x more instances in a matter of seconds. I know of no autoscaling system that could handle that gracefully, though I may be wrong and would enjoy being edified if you have a suggestion.