So, these are kinda sorta nice-to-haves. Sounds like 'imperfect' is an understatement. Essentially, there are "poles", right? In pretty much every area of polling, it's been found that societal changes of the last, say, 10-15 years have made (past) polling methods irrelevant. Possibly, these ratings poles are behind the curve in keeping up with the times(?).
I do not know what "societal changes" you refer to.
A poll or survey done with a statistical sample that is proportional on all the qualities the buyer of the data needs is just as reliable now as ever. In fact, with more core data about the population being available, it is likely that the accuracy of a poll is greater today than ever.
No matter what the characteristics of a population you want to study, it's a function of the sample recruit to find the proper target. Whether you want an accurate sample of all the significant ethnic and cultural groups in a market, or whether you want to know if households buy cake mix, the key is in sample design.
The Nielsen study has several dozen areas that are subject to proportionality control... things like the specific age ranges (6-11, 12-17, 18-24, 25-34 etc.) gender, ethnicity / race, language preference (among Hispanics), income, education, household size and the like are measured as close as possible to mirror the market as a whole with a small and affordable sample.
Nothing has happened in the last few years that changes the nature of statistical sampling and polling.
Question: Are trends of the "Trends" generally reliable?
They are "indicators" of motion. And in any case, the "trends" in the diary markets will be gone next year as Nielsen will do 12-month rolling averages with monthly issuance and full weighting for proportionality in each report. But the sample design remains essentially the same, adjusted only for continuous measurement instead of discreet "time slice" sampling.