Something I've observed as someone who works in the physical sciences and used to work as a software engineer is:

Very few software engineers understand that tolerances are fundamental.

In the physical sciences, strict equality - of actual quantities, not variables in equations - is almost never a thing. Even though an equation might show that two theoretical quantities are exactly equal, the practical fact is that every quantity begins life as a measurement, and measurements have inherent uncertainty.

Even in design work, nothing is exact. It's simply not possible. A resistor's value is subject to manufacturing tolerances, will vary with temperature, and can drift as it ages. A mechanical part will also have manufacturing tolerance, and changes size and shape with temperature, applied forces, and wear. So even if a spec sheet states an exact number, the heading or notes will tell you that this is a nominal value under specific conditions. (Those conditions are also impossible to achieve and maintain exactly for all the same reasons.)

Even the voltages that represent 0 and 1 inside a computer aren't exact. Digital parts like CPUs, GPUs, RAM, etc. specify low and high thresholds, under or over which a voltage is considered a 0 or 1.

Floating-point numbers have uses outside the physical sciences, so there's no one-size-fits-all approach to using them correctly. But if you are writing code that deals with physical quantities, making equality comparisons is almost always going to be wrong even if floating-point numbers had infinite precision and no rounding error. Physical quantities simply can't be used that way.