Answered for language learning in general. No specific language was set for this question, so the findings are drawn from general second-language research rather than one language’s literature.
The sources don't discuss "SRS backlog" or "leeches" by name, but general spaced-repetition and memory research supports a clear strategy: slow down new-card intake before your queue explodes, don't over-drill a hard card in one sitting (it barely helps long-term), and for persistently-failed cards try loosening the review interval and genuinely testing yourself rather than passively re-reading, since both under- and over-frequent review can keep a card stuck. There's also evidence that failing cards may need better encoding (context, mnemonics) rather than just more repetitions, since the review format itself matters less than whether you're actively recalling.
A large backlog usually builds up because the review schedule was too rigid or too generous with new cards relative to how well items were actually remembered — algorithms that adapt review timing to each learner's real recall probability produce far better retention than fixed heuristic schedules.1,6,8,12
In plain terms
Example
There is a sharp 'phase transition' in flashcard systems: if you add new cards faster than a certain rate, your backlog and failure rate suddenly balloon out of control, whereas staying under that rate keeps the system stable.10
In plain terms
Example
Repeatedly hammering a card in one sitting after you've already gotten it right once (overlearning) gives a short-term boost but that advantage mostly disappears within weeks — so drilling a leech over and over in a single session is an inefficient way to fix it long-term.3
Example
For language and word-learning tasks specifically, spacing out reviews doesn't necessarily help you learn a card faster than cramming it, but it does make you retain it much longer afterward — so leeches may need more spacing, not less, even though it feels slower up front.4
Retention follows an inverted-U pattern with spacing: reviewing too soon or too late both hurt retention compared to hitting a middle sweet spot, so a leech reviewed too frequently (before it's had a chance to be forgotten a little) may stay a leech simply because it's not being challenged at the right moment.4,8
In plain terms
Example
Learning material multiple times across separate spaced sessions leads to slower forgetting than repeating it many times within one massed session — this applies especially to memories that link two pieces of information together, like a foreign word and its meaning.7
Example
Studies found that the rate at which you forget something is not affected by how many times you initially studied it — meaning a leech isn't necessarily 'cursed' to be forgotten fast forever; the forgetting rate itself may be a fairly fixed property, so the fix is about when you review, not just how many times.9
Breaking study into short, frequent 'micro-lesson' sessions with built-in review helps avoid mental fatigue and supports moving information into long-term memory, which is relevant if backlog buildup is partly caused by burnout from long single sessions.2
In one study, students who did any kind of repeated review activity (lecture, e-learning module, or hands-on small-group review) retained far more long-term than students who did no repetition at all — but the different review formats performed about the same as each other.5
Flashcard and spaced-repetition tools only work as intended when the learner actively tries to recall the answer before checking it; simply reading over cards as notes (passive review) weakens both the 'testing effect' and the spacing effect that make the method work, which may explain why some cards keep resurfacing as leeches.11,13
In plain terms
Example