1. Use a TripleByte replacement (e.g. https://www.otherbranch.com/) to filter out obviously bad applicants. Basically, job-seekers do a long set of interviews, and if they pass, are considered generally competent.

2. If you get a lot of generally-competent employees after applying reasonable filters (e.g. matching skillset, expected salary), don't give them a long automated test, pick a smaller set randomly. All of them have demonstrated competence, and the likelihood that the test will give you more the more competent employees is offset by the likelihood that they'll move forward with applications more respectful of their time.

3. Do final-stage (human) interviews with the small set of employees, where you test specific skills relevant to the job. Here you can also throw a couple general-skill questions to ensure the applicant really is generally competent; it's not disrespecting their time, because it's part of the interview time and you're spending it as well (maybe it is if the entire interview is especially long, but then you're wasting also your own time).

The important part is 1). Otherbranch may not be good or popular, but at least if/when employee supply falls below demand, "mass interview" seems like something employers will need to filter out bad applicants without wasting good applicants' time.