Topical and special issues are not uncommon and can be a valuable addition to the scientific literature. I know about numerous cases where selected participants of a conference were asked to contribute to a special issue in a particular journal on the topic of the conference. This is not unlike conference proceedings, but usually with a higher quality level, because the contributors were carefully selected. It also happens that the editors of a well-established journal invite contributions on a special topic. There is nothing basically wrong with special issues per se.
Now it appears that publishing special issues is part of the business strategy of a number of publishers such as MDPI, Hindawi and Frontiers. Judging from the number of unsolicited invitations I get, often hardly related to my domain of expertise, it appears that these publishers leave out the “carefully selected” part. Why would they do such a thing? I think that the answer is very simple: special issues allow them to delegate a large amount of work to the guest editor. Instead of having to solicit articles themselves and handle their peer review, the guest editor will take care of all this work, resulting in a non-negligible number of papers, which, in the gold OA model, means that more income for the publisher.
Why would researchers agree to participate in such a scheme? There are probably a number of different reason. Some feel flattered to have been asked, as it seems to confirm their importance in the field. Some may be genuinely interested in providing a topical review of their domain. Others accept because a controversy in their field opposes them to some powerful rival, whose influence they hope to circumvent.
In the end, some of these special issues may turn out to be very good, if the guest editor does a serious job. So where is the problem? Mainly, quantity rarely goes with quality. The business plan of MDPI and others appears to consist in contacting large numbers of potential guest editors and providing huge computer-generated lists of potential contributors, a process I am inclined to call Monte Carlo publishing. The law of large numbers implies that such a broadly scoped process is bound to produce a few very good issues, but at the price of also producing a large number of mediocre ones. As a collateral damage, the reputation of well-meaning researchers who published in these issues gets hurt.
While this Monte Carlo strategy seems to generate a substantial short-term income, I wonder whether it will work out in the long run. Many colleagues are, like me, annoyed by the large amount of spam invitations in our emails, and certainly would never submit their research to publishers relying on such a strategy. There have been other controversies, like this one where Frontiers was accused by a number of their board numbers to favour quantity over quality: http://news.sciencemag.org/people-events/2015/05/open-access-publisher-sacks-31-editors-amid-fierce-row-over-independence
If as a researcher you want to preserve your good reputation, it is important that you examine carefully with which publisher you get involved, be it as an author, an editor or a reviewer.