IssueTracker

    assertTrue(Double.NaN.coerceIn(0.0, 1.0).isNaN())
    assertTrue(Double.NaN.coerceIn(0.0..1.0).isNaN())

https://github.com/JetBrains/kotlin/blob/master/libraries/stdlib/test/ranges/CoercionTest.kt

Low-hanging fruit. Shalom, mind taking a look during your work for PIs and loaders?

We can check for NaN, but what is the expected behavior here if a NaN is passed in?

1. We can throw an exception, but that would end in the same result you are experiencing now (i.e. a crash if you don't handle it)
2. We can just set the progress to zero in cases of NaNs, but then you would not be able to tell that you have a problem in your implementation and you are passing in illegal values.
3. We can document it that we only expect a non NaN values, but that seems a bit redundant and we don't usually adding these to the docs as it's implicit in most cases.

On my end, I discovered it was caused by division by zero and just set the bar to indeterminate in that case.