In 5 years the design flaw pointed out by Microsoft research can become relevant.
If you can't remember what that was about:
Without a sufficient block reward nodes will stop broadcasting transactions because by not doing it they increase their chance of getting the fees contained in them if they are mining.
Today, the huge majority of nodes are not mining.
What do you think the ratio of pools (mining nodes) to nodes (non-mining) will be in 5 years?
I doubt it will be very much different than it is today.
I also doubt the block chain will be prohibitively (as in too costly to be a node) large in 5 more years. I can still keep the entire thing in RAM if I wanted too.
A 12.5 bitcoin block reward isn't sufficient?
If that actually is a problem, I think it will take much longer than 5 years to become relevant. If light nodes still broadcast transactions, it's probably never going to become a problem in the first place.
If the trend holds it might be at 6.25. The transaction fees might make the difference between break-even and profit. What you are saying the problem will not arise because of altruism (which is running a node without a financial incentive)
People running a node because they provide some sort of service and do not mine might want to sell their transaction data.
Read the paper a solution to this problem along these lines is actually proposed in there. When it was published it was harshly criticized but maybe people have become more receptive.