An agent has to be physically present to affect it, first of all. This helps a lot by constraining the field of actors from "potentially anyone with an internet connection" to "people who physically come in contact with the artifact". This is a widely underrated safety benefit with applicability all over, not just in elections.
The great benefit of paper voting is its simplicity. Any time you invent a complex contraption and tell the people to just trust that contraption's designers, e.g., the "blockchain technologists" you've mentioned, you've just built a new ruling class.
For something to work, it must use a process and mechanism that is plainly and obviously valid to anyone of sound mind. "Counting marks on paper" is such a process. You can't get much more complex than that before you're transferring control to a new elite.
People will always search for ways to unduly influence and damage processes by which power is ascribed. This is one of the main reasons why a process with trivial verification mechanisms is required for the public to credit the results.
In the year of Spectre and Meltdown, it is immensely naive to pretend that computers are ultimately trustworthy. Simplicity remains supreme, and computers are not simple.