I also have a 2 year old here with these (imagine my surprise to see this on HN), and I've troubleshooted more than one track creation. I can confirm the findings of the above poster. They don't buckle upwards much. There's some margin for error in the connectors that allows for the tracks to pivot some. A degree or two off and you can still get the connector to fit, but you'll feel the tension in the track as one side is fitting much more tightly than the other due to the bend. So introducing another track segment somewhere in the loop (the link goes into the math behind this, but a little observation and intuition will also yield the correct result) will ease the pressure. In my experience this is almost always caused by trying to close the loop a little too tightly.

Edit: Re-reading the rest of the "look for track flatness" comment; the second and third sentences about tolerances and bowed joints are spot on. For example, looking at the final track layout for the "mathematical" approach, I can tell you that I'd have no problem shifting that track down an inch and snapping it in place.

Duplo's were the go-to toy in my house for years. The larger size makes it much easier to find pieces in "the big box of Lego" than standard Lego's. Duplo and Lego, in general, have amazing longevity — they were the best toy investment we made over the years. :-).

As an aside, these articles are the gems that keep me coming back to HN.

I'm really curious now. I haven't had 2 year olds for a while. Can you try this and see? Surely there is at least enough warping that you'll see a 1mm rise?