The problem with Kremlinology is roughly the same as what you get with the study of similarly restricted states - and the more restrictive the state is, the tougher it is to conduct meaningful analysis. Simply put, conventional sources of information are often unavailable or are otherwise difficult to access.

As you say, sometimes it devolves into making judgments based on otherwise inconsequential information like what people are wearing on official photographs, and whether or not some people are present in them (or where they happen to be standing).

In general, I would say that there is a definitely a difference between "Kremlinology" and genuine scholarly research on the topic. Not to say that the Kremlinologists never got things right, but the methods and sources that were available to them were very limited. In the nutshell, classic "Kremlinology" implies subjecting the Soviet press and other published material to thorough examination, attempting to ascertain and judge Soviet policies and, importantly, predict Soviet actions (something that a relatively small proportion of purely scholarly material sets out to accomplish). Naturally, the biggest problem here is the trustworthiness of the published materials in question - at worst they contain completely fabricated information, at best they leave many points uncertain and under-illuminated.

So, working from there I would also say that the viability of "Kremlinology" depended largely on when we're talking - specifically the degree to which the Soviet Union was closed to outside observers. Simply put, when you can just go to Moscow and do research there, even with some limitations, you obviously are better off doing that instead of trying to analyze state-sanctioned Soviet press sources. In other words, Kremlinology was more of a necessity at times, brought upon by the lack of viable alternatives. Especially during the Stalin period the researchers had little to go by apart from the Soviet media. Ironically, at the time the same can largely be said about the general Soviet populace.

Arguably, the biggest blow to the credibility of classic Kremlinology (or the state it ended up evolving into by the end of the 1980s - more on that later) was its failure to predict the collapse of the Soviet Union. In fairness, the same can be said of pretty much everyone else who was studying the Soviet system at the time of the collapse, including the more traditional academic research - but then again, most did not set out to plot USSR's course for the years to come as their ultimate goal. With that in mind, a lot of the Kremlinologists openly admitted that their approach offered too little reliable information to properly understand what is actually happening in the country, and a lot of them reverted to some of the more conventional research methods when and if the opportunity presented itself to them.

The proportion of Kremlinology that tended to end up on the pages of regular press tended to be almost entirely speculative in nature as a result, and hardly anything of what it "predicted" was presented in concrete terms. On a personal note, I can say that much the same thing can be said about the contemporary reporting of mainstream press on Russia - they rarely refer to actual specialist on the matter, that is to say the people who studied Russia professionally, or were otherwise engaged with Russia (and IN Russia) in some professional capacity, and rely on either anonymous sources, or sources with shaky credentials - for example this article from The New Yorker quotes Andrei Kozyrev, the foreign minister between 1990 and 1996, but hasn't been active in politics since the late 90s, and since 2012 has been living in the States, while Oleg Kalugin, a former K.G.B. general they also referred to, has been living in the States since 1995, so their viability as reliable sources of information for analyzing contemporary Russian policies is very questionable. The situation was much the same with most of the Kremlinology during the Cold War. So a lot of the "predictions" were more semi-educated guesses, and a lot were purposefully ambiguous in their prognosis, so it is hard to accurately gauge their predictions. Those that weren't ambiguous were instead overtly ambitious - claiming that they could ascertain that the leadership struggle was becoming more heated because the portraits of Beria and Malenkov changed their spots twice within the same day during the October Revolution Day festivities in 1947 for example. At the time, when there was nothing else to go on, these predictions were literally everything you could work with, and were treated very seriously, but in retrospect their conclusions were more often off the mark than not.

As the conditions for reporting on Soviet Union loosened up after the death of Stalin, serious Kremlinology started to fade away, becoming history's equivalent of alchemy so to speak. By the mid 1970s it was mostly an anachronism and was already largely used in a pejorative sense to describe the "guessers".

Nowadays the reasons for employing the same methods, that is to say Kremlinology in relation to states such as North Korea, are the same as before (with the same implications for producing accurate analyses and predictions) - the lack of other available information.

Hopefully this answers your question, at least in part.

Some decent relatively recent academic articles on the subject (behind the paywall unless you are accessing them from an academic institution I'm afraid):

• Zachary J. Jacobson. "On the ‘arcane modern science of Kremlinology’ or the case of the vanishing birthdays," Cold War History 16, no. 2 (2016): 141-158.

• Stephen Kotkin. “The State – Is it Us? Memoirs, archives, and Kremlinologists,” Russian Review 61, no. 1 (2002): 48–49.

As a follow up-question, did the Soviets have an equivalent (as in "Washingtonology") and if so, did it provide any useful information for Soviet intelligence?