But what is the purpose of keeping tokens without jobs? For example, we have infinite TTL, cause we waiting for other services to up. And we can't remove tokens at all. May be an additional API function for removing it will do the trick. We just don't want to have "dead" values in Redis forever. If you agree, I can prepare pr for a new function or for removing tokens inside clear_* and remove_* function families.

But what is the purpose of keeping tokens without jobs?

There might be some weird cases, but I generally agree with you. When a user deletes a job, they might expect the token to be deleted as well. Adding a separate API to clear tokens would be overkill; it can be an option in the same function and default to true.

Please feel free to open a PR.

By the way, I am also not sure it's possible to delete all tokens since some tokens might be supplied by the user as well (not derived from job args). In that case, it's okay not to support those. Thinking again, it might be better to keep the default as false since there is no straightforward way to figure out if the job has unique tokens attached without doing additional redis operations.