Worse yet is that there is a world of difference between being able to recognize a word when primed, and being able to recognize it when not primed. When it decided that I need refreshing on a phrase I probably got flashcards corresponding to translate this English phrase into Russian, translate the Russian phrase into English (from a set of pre-chosen words), type this Russian phrase in Russian, and maybe type the English version of this Russian phrase and/or type the Russian version of this English phrase. All with the same phrase. Lumped with many other variations of the exercise on the same topic.

Introducing them all together is helpful. However after some level of proficiency I would like to receive them separately. In a group with unrelated exercises. On a schedule indicating how well I know that phrase.

That said, they have been a painless way to gain a basic foundation.

My observation from using Duolingo (Spanish) was similar to this. My theory was that there spaced repetition was simply based topic and then once a topic was chosen it then decided what words to test you on. This is crude, simplistic and ineffective (IMHO). Like you say, there are words you see rarely.

Also, certain topics end up reinforcing earlier topics. I lost count of the number of times that the stupid "algorithm" would make me go through the exercise of translating "no, nada". At some point, don't I just know this?

Additionally, when you have topics that aren't "gold", it simply selects the first non-gold one to repeat. This can mean that a bunch of stuff at the bottom of the tree gets rusty because you don't get to it while you end up redoing stuff you know backwards.

I made this observation in the forums but couldn't really get any traction on the issue.

At some point they also introduced the stupid heart system, which was pretty much my trigger for just abandoning the whole thing.

Also, I'm really not sure how much this format actually works. Like having completed the Spanish tree, I'm honestly not sure if I really remember anything more than a small fraction of it.

Perhaps that's because they have too many words?

If you look at some of the core languages like German the course has a 'words' tab and it shows you exactly how strong it thinks you know each word and when you saw each word last.

It could be different between languages, and who knows how well it works.. but it's definitely something they seem to be working on and putting thought into.

I've found that Lingvist is pretty good at spaced repetition. It focuses on hammering in the words I don't pick up quickly, but it still reviews all of the vocabulary it offers pretty regularly.