When those assumptions break, accessibility issues start showing up very quickly.

Having said that, yes, tending to the needs of disabled people absolutely does help identify design shortcomings and making software better for disabled people often benefits far more users than than the targeted group. This is called the "curb cut effect" or simply "curb cuts" in the industry because like sidewalk curb cuts made for wheel chairs, the improvement was actually a win for everyone from jogger to parents with strollers, to rolling luggage, delivery people with hand trucks, etc. etc. When we make things better for one group, often many groups benefit so designing with everyone in mind, and not just people like you, is always worthwhile.

While thinking about this, I realized that accessibility often acts like a kind of stress test for design assumptions. If something only works under ideal conditions, accessibility issues tend to surface those weaknesses very quickly. (I wish I'd known that when I was a tester!)

The best lesson in accessibility to learn is that our societal needs are complex and the various standards exist for good reason. If you want to create a complex and particular design using CSS that is fine but keep the tagging underlying that design compatible with screen-readers and allow easy overriding of styling.

One of the most frustrating things for accessibility is advertising since it specifically goes to lengths to use obfuscated class names (to avoid ad-blocks) and bright colors (often via images/videos that contain embedded text). At some point I really do hope we realize just how expensive advertising is and how many externalized costs it forces on us all.