How do you design a system that is barely used most of the year and then experiences one huge spike only one day? Do they have tons of capacity sitting around most of the year?
Schedule your hardware upgrade cycle and lower priority workloads around the event so that you can get your hands on several times more hardware for a month or two.
</tangent>
How big is a tax return? Some XML 10k/100k in size?
What do you need to do: save the file, maybe do some basic sanity checking and store the SSN of a person on a db.
It is complicated, but not impossible.
Obviously you can scale systems ahead of time if you know, but you still run into issues because you probably haven't scaled the entire system and don't know the likely bottlenecks since it's hard to generate that kind of traffic in a QA environment.
Think of the scale of facebook, and then realize most of their payloads are <1k json/protobuf. It's not the size of the payload, it's the delta in quantity.
Note: just about every online store must do the same type of thing around specific holidays.
[0] - https://federalnewsradio.com/tom-temin-commentary/2018/01/ir...