jerf described a similar case where one is okay with losing some data, but not "too much". I'm not sure how to make that approach rigorous. (If rigor isn't important, and it's just best effort with no guarantees at all, then don't bother with fsync at all.)

I'm also not sure in what higher-level use case it actually makes sense. Sorry if I'm being thick, but saying "the case where I want exactly that" doesn't help explain that case :) I'm looking for a higher-level description of the problem that would solve (e.g., "an ACID database", or "a CDN" -- except those are cases where you may want zero or any data loss, respectively).

In most things I do on my computer, a few seconds or a couple minutes of data loss is acceptable, but data corruption is not acceptable. Without ordering guarantees data can become corrupt, something from weeks ago can disappear because you tried to update it. that is what is unacceptable.

Perhaps compare to the uberblocks on zfs. You can wipe out the most recent one, or the most recent fifty, and still have a consistent, slightly rewound, filesystem. Take that level of consistency and add a notification when everything is written out, and you have a nice nonblocking fsync.

The kernel might wait for a very long time before writing any data to disk (especially in laptop mode with a spinning disk drive). Telling it to start writing now seems like a useful operation for an application that can handle the loss of recent writes, but would prefer to have data start going to disk anyway. Otherwise a power failure could mean losing 30 seconds of writes or more.