Actually Spotify has had advanced search syntax at least since 2011: http://web.archive.org/web/20110718125428/http://www.spotify.... It just apparently is not advertised by them particularly well to the users. To search from label Decca, just search label:decca (and then select, for example, "See all albums"), or add label:decca to your more specific search if you want to use it as an additional filter. To be able to tell which of the "Requiem in D Minor, K.626:3. Sequentia:" is "Confutatis" — depends on which device you use. On my computer when I select "See all" next to "Songs" in the results, then the track title column is wide enough to show also "Confutatis", or it is also possible to right-click on a track there and select "Show Credits". On a phone one option would be to add "Confutatis" to the search term, or you could start playing one of the results and then you can scroll left-right with the finger on the currently-playing-info-bar below and see the full track names there.
The author's conference paper [0] has more details, although not enough to remove all the doubts.
But I guess the main goal of the article is to show that, even though "radiation detector" sounds like a complicated thing to most people, or at least requiring some nontrivial components, it is actually possible to get started with DIY experimentation with very few and simple components and little effort. You might indeed get better results with ZnS:Ag phosphor sheets, but by showing that it is possible to build at least something with just a webcam that people already have you massively increase the likelihood that the reader actually starts to build and experiment. And when people already start experimenting, then many of these potential problems get incrementally noticed, figured out, and removed, and all kinds of cool projects can result. But yes, it would, indeed, have been nice if the article would have provided more information about testing the detector against various potential interferences and problems.
I guess the author was thinking about instrument accuracy, which is usually expressed as percentage of full scale, or percentage of the actual reading, or percentage of the scale span. So I would guess the author wanted to say that it would take the detector about 20 hours to reach the accuracy of +/- 10 percent of the reading.
In terms of magnification, yes, but there are also many other variables that can be relevant, so I guess it's good to have multiple cheap designs to choose from, depending on the goal and circumstances. For example Foldscope seems to be using ball lens, which indeed gives higher magnification, but probably also suffers more from aberration, may be trickier to fix firmly and precisely into the microscope housing compared to the lens that are already created on a flat plate, etc., so for those tasks where lower magnification is sufficient or preferred the droplet lens might sometimes be more suitable.
There's quite a nice and easy to use program (a programming / modeling environment) for exploring and getting an intuition of how a system of many individuals who follow simple rules and interact with each other can give rise to interesting complex behavior. The modeling environment is called NetLogo, and it's basically a multi-turtle version of the old famous Logo programming language. It has lots of different models included in its example model library, including about ants, and the models have modifiable source code. I have quite enjoyed playing around with it, and have even used it in some presentations about how simple rules can give rise to complex behavior.