You should be making your systems resilient to traffic spikes (being fault tolerant helps) and auto-scalable. That will make sure that the impact of the spike is mitigated. Technologies like Kubernetes that we are using in production help us in that also. If instead of a spike you have a continuous traffic increase you should be looking to optimize and continuously improve the efficiency of your code.
All your scaling principles should remain the same, the mindset you apply, while building, is what changes when you move to Elixir.