Or put another way: if voter ID were considered a serious issue, then the obvious first issue to solve would be establishing a new national ID system.

This would solve other problems - for example the use of insecure social security numbers as de facto national ID, and passwords, which you give to everybody yet can be used to steal your identity simply by knowing. Like, this is actually a big problem in the US that affects numerous people, and would save the economy million of dollars minimum, probably more, by solving in a secure way.

Then, with reasonable confidence that national ID had the uptake rate necessary to pull it's lack of below the level of impact which would effect elections, you could make voter ID mandatory.

Weirdly, no one ever seems to start with the - far more problematic - "hey national ID is kind of a big problem..." when they propose voter ID though. They also don't start with any data about in-person voter fraud (except if you count all the people who get caught trying to commit it to "prove" it's totally easy to do everytime this comes up).