[Malte on discord]

I think allowing infinity for weights makes sense in general, especially in combination with evaluators like pref() that only return 0 or 1.

I think this an artifact of disallowing negative values. To disallow negative values using a "Bounds" object, you also have to set an upper bound, which will then be shown explicitly. If we changed other evaluators to rule out negative values using Bounds, they would also say "up to infinity".
We can discuss changes to the behaviour of Bounds objects if we don't like this.

The weight for the weighted evaluator is documented to be an integer.
However, the parser accepts 'infinity' as the weight parameter.

The way we parse int arguments blurs what is meant to be the largest number the hardware allows and what is meant to be the symbol infinity.
For example the random_seed is bounded by [-1,infinity] here we do not mean the infinity symbol but the largest number the hardware allows.
For the ConstEvaluator it is bounded by [0,infinity] but here we mean the symbol infinity to get the constant deadend evaluator.