I think dolly's complaints are a bit ironic really: the vast majority of applicants for maths at Cambridge are invited for interview, precisely so that they have the opportunity to demonstrate their potential. Everybody who is interviewed has a realistic chance of an offer, based on their UCAS application. Even if they don't get an offer, it is not a waste of time as the whole interview experience is very useful in itself for a prospective maths undergraduate.
One could set filters more harshly and invite less applicants to interview, but most filters would discriminate against students from lower achieving schools or from backgrounds with no tradition of university study. For example, pushing up the cutoff for AS UMS tends to favour students from selective schools.
Interviewers need to be aware of this and they need to give every candidate the opportunity to show off what they can do rather than sticking with easy questions.
Why on earth would you think interviewers are not aware of this? Aren't you assuming something about their own backgrounds in thinking that they aren't aware of the costs of travelling?
The only reason an interviewer sticks with (relatively) easy questions is if a student cannot get very far with any of them. What on earth would be the point of asking harder and harder questions, tearing the applicant to pieces, making them feel more and more stressed? Isn't it far less cruel to stick with a pleasant discussion rather than pushing the candidate until they collapse?
maybe the Camb interviews are more to see how they would fare in tutorials, as they know the hard questions will be faced in STEP anyway?
The (decision making parts of) Cambridge maths interviews are basically STEP questions but it is considered hard to distinguish definitively between candidates based on short interviews/short tests before interview, hence the over-offering based on STEP grades. This is anticipated to change with the removal of AS: definitive offers are likely to be made based on longer pretests akin to Oxford's MAT. Imo the selection of maths students by both Oxford and Cambridge inevitably has a certain amount of error: the top candidates are obvious but for many others you are looking at very small differences and trying to correct for differing backgrounds.