Music tests (which are not "surveys") can cover as many as 1,200 songs for deep library formats like Jack. Many will not "pass the test" but no programmer is discarding songs based on personal preferences.
Consultants don't do music tests. They may recommend a research company, but they don't do the test. They may conference with the PD and station's company format specialist or national PD, but there is no place for "skewing" test results song by song.
Yes, that happens. Songs that were huge "back in the day" can be hot, medium or cold today. There is no predictability other than asking listeners a couple of times a year how much they want to hear the song today.
A "focus group" is not used to test music. A focus group is about 8 to 12 people who chat with a moderator about things like station perception, things they like or dislike in the morning show, and so on.
A music test, mostly done online, has about 100 people who are professionally recruited score "hooks" of between 300 and over 1,000 songs as to how much they'd like to hear the song today.
Recruiting is done by research recruiters who work locally to find people that the station and the research company have decided they want to have in the project. Generally, they are heavier station listeners, and inside the core target demos. They are balanced within that group for demographic, gender and even ethnic groups.
Depending on the market, a recruiter will be paid around $100 to $150 per person for recruiting, and each participant may get $80 to $200 to score the songs. A total test can cost $20,000 up to $35,000.
If you are in, let's say, LA, with 10 million people over age 18, a station may do a music test twice a year with 100 people each. You do the math. Your chances of being recruited by any station that does local testing are about once every 80 years or so.
By the way, I've done huge tests of over 400 people, and then extracted many sets of 100 at random. Once you get over 80 correctly recruited people, there is no gain in accuracy from adding more respondents as long as the recruiting is done correctly. But that 400 person test costs four times as much... almost $100,000 and no station can afford that if 100 or less yields the same results.