An alternative explanation is that we have a lot of "backup capacity" such that small amounts of damage are not noticed by reasonable tests (or potentially are impossible to measure) until we reach a point where we fail to function. I have enough beta cells in my pancreas that you can destroy a lot of them and you will not see my blood sugar lose control; maybe it takes slightly longer to steady, but unless you have previous tests of how well I function I might still look "normal", but if you destroy enough then eventually I stop being able to regulate blood sugar and I clearly have diabetes.

Maybe if you kept tons of test results of when I was normal, you could try to compare, but in addition to the problem that most people don't take tests when they don't notice issues (and even when they do, the medical profession just seems to love to destroy files after a number of years: it just feels criminal to destroy data like that on someone's medical history), the test subject is also getting older and you expect them to not function as well :/.

Even if you know something is failing, often medical science just doesn't care: over a decade ago I was getting hearing tests done occasionally as I could tell I was losing the hearing in my right year, but I had hearing that was so good (better than most people even for being in my early 20s, when everyone still has good hearing... and so usually aren't taking hearing tests) that the doctors would explain I'd topped out their charts, had "excellent hearing", and they couldn't even know if something was wrong. Well: I definitely have extremely worse hearing in my right ear now :/.

It wouldn't surprise me if the exact opposite were true: that medical science often estimates things assuming there are "thresholds", when quite possibly almost nothing is subject to a threshold effect while often being cumulative; the issue being just that the response curve is sufficiently sloped, and is further skewed by tests based on results instead of on internal state, that it looks like there are thresholds, and so we underestimate essentially everything dangerous.