Some cards I have failed twenty times and I keep dutifully reviewing them, when should a card just go?

The flag is a message and you have been reading it as a compliment.

A card you have failed many times is almost never a hard fact. It is nearly always a broken card, and the usual faults are:

It asks for too much. One card with three things on it fails if any one is missing. Split it into three.

It is ambiguous. More than one answer would be correct, so you are guessing which one you wrote. You fail on the wording, not the knowledge.

It tests recognition rather than recall. You know it when you see it and cannot produce it, so it feels known and scores as failed.

It has no context. An isolated fact with nothing to attach to. Almost impossible to retain and this is the commonest cause.

You never actually learned it. Repetition maintains knowledge; it does not create it. If you never understood the underlying thing, no amount of reviewing installs it - this is the one people most need to hear.

So the response to a repeatedly failed card is not more repetitions. It is to look at the card and diagnose it. In my experience the majority are fixable in two minutes, and the fix removes them from the problem pile permanently.

The procedure I use when a card gets flagged, in order:

Ask whether I actually need this. Honestly. A lot of cards were added in an enthusiastic first month and correspond to nothing I care about now. If the answer is no, delete it and feel nothing - see below.

Suspend it and come back. Sometimes the problem is that it depends on something else I have not learned yet. Suspending and re-enabling later works surprisingly often.

Rewrite it. Split it if it has multiple parts. Add a clue that disambiguates. Add context: where it comes from, what it relates to, a mnemonic, an image. Turn a recognition card into a production one.

Learn the underlying thing properly, away from the deck, then rewrite the card.

Then delete, if none of that helped.

One caution: rewriting resets its history, which is correct: it is a new card and its old failures were the old card's fault.

And a preventative note for new cards: the single most useful rule is one fact per card, phrased so there is exactly one correct answer. Most leeches are created at the moment of writing, not discovered later.

On whether deleting undermines the system: it does the opposite, and this is worth being direct about.

The deck is a tool, not a record of your intentions. A card you keep failing costs you time every session and teaches you nothing. Multiply that by a handful of cards and it is a meaningful share of your reviews spent on the material with the worst return.

Deleting it is not giving up on knowing the thing. It is declining to use the wrong tool for it. Some things are learned by use, by reading them in context, by doing the task - not by a flashcard.

What I have found after deleting freely for years: nothing bad happens. I have never once missed a card I deleted. If something genuinely matters it comes back, in context, and I add it again in a better form.

A useful periodic habit alongside it: go through your deck occasionally and delete anything you no longer care about, not only the failures. Decks accumulate the interests of previous versions of you, and a smaller deck you review properly beats a large one you dread.

And on the specific handful eating your sessions: fix or delete all of them this week. The change in how the deck feels is immediate and out of proportion to the number of cards involved.