That's what I was thinking. Don't need fancy graphics, just make the back button work further back than the tab creation. Doesn't even have to be a copy, if they make the history a tree structure in memory.

That last part doesn't work, unfortunately. We looked into doing exactly that, but the way history navigation works is specified quite strictly - and relied upon by existing web content - making this infeasible.

I don't see how this could be true. From the perspective of a web page, it'd be exactly the same as if the user's new-tab clicks were regular clicks instead.

From a user perspective, my main pain point is the prior history running out at the point where I opened a new tab. In many cases I've already closed the tab it came from, and if it's not closed I have to go looking for it.

This seems like the most simple solution.

Are there any disadvantages to this approach?

See my reply in another conversation thread: https://news.ycombinator.com/item?id=13520978

To a degree that is what we are doing here. There are some technical difficulties there and we plan to write followup post on that as well. That being said having a visual representation of the navigation trail is very useful even if it's just to gain ability to jump few steps back rather than going step by step.