The API Reference of bindAggregation explains that the binding info property key is useful to improve update behavior in models.
The name of the key property. [...] This can be used to improve update behaviour in models, where a key is not already available.
However, that explanation is somewhat incomplete which makes me wonder in which cases it's actually useful to provide a key property.
Sure, I can provide the key on every single aggregation binding definition but then I'm doing something without knowing "why" exactly. On the other hand, the above quote also tells me implicitly that there are cases where a key is already available. So I browsed through the source code and found that ODataListBinding (V2) has getEntryKey which provides each binding path as a key value. However, that method is called nowhere in my projects.
Has anyone experienced any significant improvements of "update behavior in models" due to the provided key?
I'm aware of this case where a key is helpful to allow combining two-way data binding with growing enabled controls (despite the doc saying that it's not allowed to combine them together). But I don't think that's the actual use case of key.