I think my views here might be biased since the only people I see interviewing in Python tend to be bootcamp graduates or new grads, and their coding performance in general tends to be not as good (which is fine, as the expectations for L3 SWE are lower).

I agree with you that the Python language syntax is nice, but it doesn't have quite the same level of built-in support for algorithmically useful data structures as C# or Java with Guava. To give you a concrete example, it's not unusual for interview questions to require the use of a self-balancing binary search tree or a similar data structure in order to reach optimal runtime complexity. Realistically no one is ever going to implement a self-balancing BST along with solving the actual problem inside of a 45 minute coding interview, but it's nicer if you're able to refer to an actual library implementation that exists and can be used versus hand-waving away the existence of one.