I never worked as a trader, which means I cannot answer the specific question. However, I have seen (from both sides) plenty of similar applications like HackerRank, Leetcode and the like. Personally, my experience makes me question the usefulness of such tests in finding valuable candidates.
That said, I think these tests are quite useful in getting rid of false positives (candidates one thinks are good but turn out unreliable in the end). Although I think it is at the expense of the candidates who have to study something that is often rather useless for their job. These test may also create plenty of false negatives (eliminating very good candidates) but if there are enough applicants, this is not a big issue (for the hiring firm). The good thing is that the exceptionally gifted candidates (which I am certainly not) will almost surely pass such tests as well.
If I can quickly find all strings formed from characters mapped to digits of a number, or correctly return an array of all the palindrome pairs of words or write all sorts of linked lists is more often than not just hard practice.
Prior to COVID (currently one needs to be happy to find job applicants across the spectrum in my experience), there were a lot of applicants for few available jobs. Using screening methods will leave only a few potential candidates. In reality, plenty of companies use computerized HR tools to pre-screen already before that (with thresholds for degrees, languages, experience etc). If your application doesn't tick enough boxes, you never actually get in touch with a person. Insofar, these tests are in designed to get rid of false positives.
The downside is that these tests tell you very little about the actual performance of a candidate. The scenario that @amdopt points out will involve real stress. If it were a simple multiplication within some range, there would be no need for an experienced partner to step in.
A pass mark could simply mean a candidate spent a lot of time studying these types of questions. While one might say that being dedicated is an important trait, I think these tests still tell you little about the actual performance in a work environment. If you have done the CFA or something similar, you will probably have realized that lots of people with little actual, applicable knowledge can pass these exams, provided they studied a lot (of practive questions). If one robotically memorizes these questions (or practices multiplying random numbers in her head for weeks), it is indeed a waste of time and resources, although it may actually end up helping you find a job. However, that time could have been spent more wisely (in my humble opinion).
The CRO at the bank I currently work at once told me after the first training sessions I did with a new employee, that he never really cares much about the immediate performance. In his opinion, one can only tell after a year or two if someone has potential or not, which by his definition means that one can independently carry out a reasonable amount of work, with the right amount of detail and without too many errors. In the words of Nick Patterson, “Do you notice when your results are obviously rubbish?”
It is one thing to be able to explain how a digital is replicated with a tight spread, or to code a quick app doing this (the interview type question), and another to realize if the value you get makes sense (if your inputs are sound...).