You keep believing that. Data integrity is no different now than it was five, seven or ten years ago. Not "crashing" is NOT a determining factor for stability in 15 minute runs. 15 minute runs are ONLY good for checking thermal compliance.
This is known, by virtually every tenured overclocker or serious enthusiast. 15 minutes can't tell you ANYTHING about your stability. With only that amount of time invested in checking the stability your are almost certainly introducing micro-errors and silent data corruption into anything processed and hopefully not of it is mission critical or cumulative in nature.
Silent data corruption due to unstable, or too far out of spec to be acceptable stability, is not new, or old, it just IS. There are no programs I nor any of the other people I talk with who have even more years of experience than I do, and I've been at this since about 1982, know of that can determine ACTUAL stability or anything close to it without at least minimal long term testing. 24hrs is still the standard.
For gaming systems, it is unimportant. You are not worried about cumulative or silent corruption of data because there is rarely anything of long term importance being written or re-written, so it's largely irrelevant. However, it IS a factor for the operating system and supporting files on those types of systems, and for non-gaming work or productivity related applications and data it is a major concern.
There has been clear evidence showing that micro-errors and silent data corruption can account for much of what gets reported as software bugs, faulty device drivers and file or operating system corruption, even on systems that are not overclocked, but at far higher rates on systems that are overclocked without being properly checked using conventional stability testing mechanisms.
Averages are one in every 1016 bits are corrupted on an average test of an stock clocked or stability verified system. On unverified overclocked systems those rates have been shown to be four to twenty times higher. That's fairly recent data too, not just some mythical commentary from twenty years ago.
https://www.microsoft.com/en-us/research/wp-content/uploads/2016/02/eurosys84-nightingale.pdf
So, sir, I beg to differ, and factually, I am NOT wrong.