In my view this is a poor article and may provide more illustration of the limitation of the El Reg mind. If a site is coming up on the planned spend, what is a company to do? Continue spending or cut back? It seems the author's view is that a company should just continue spending.
Maybe that's the right course for some. But AWS is pointing out a way to contain costs is to stop using multiple sites behind a load balancer and instead drop back to one server without a load balancer.
If you are already running multiple web sites behind a load balancer - as many are - then you are already running multiple web sites. AWS tools like cloud formation and containers make it straight forward to add and remove servers based on demand. Normally only the internet facing load balancer needs a static IP address.
But the premise behind the sarcasm in the article is misplaced because it is not true to say that it is necessary for two 'versions' of a site to be maintained. I think this demonstrates the author doesn't really have a great understanding of AWS or perhaps, wrote the article with impaired faculties.
AWS is pointing out that using a lambda function, possibly triggered by a billing event, a company could call a task that might:
1) Remove the static elastic IP address from the load balancer;
2) Shutdown the load balancer and all but one of the EC2 instances hosting the web site; and
3) Associate static the IP address with the remaining web site EC2 instance.
Job done. The web site is now using one machine and cutting back on costs. Is it difficult? No. Does it require the maintenance of a special version? No.
It's not an elegant approach to cost management but it seems likely the comment seized upon by the author was an attempt to remind users of the options they have available to control costs automatically not a statement of best practice.