Demonstrate with data that moving that button 3px to the left will increase sales by X% or customer retention by Y% or whatever our business metrics are.
There's a difference between "obsessing over the details" and "tweaking for tweaking's sake". Make sure you're really doing the former and that the particular details you're obsessing over matter measurably.
I envy everyone working in natural sciences. Their jobs are immensely challenging, no doubt, but at least they have a clear path forward, more or less. But they are making impressive progress all the time.
Looking at human behaviour is much more challenging (so challenging that hardly anyone has an idea how to do it and most stick with boring unreliable ways that have somewhat kind of worked in the past), fraught with ethical issues, and so on.
Measuring everything is not a realistic goal. Measuring should be done whenever possible and is an awesome tool every designer should be using but it is impossible to employ it for every little change. And it has so far not been possible to develop a general theory of design that would help you decide small things without testing that specific instance of the design.
What frustrate developers/engineers are their code changes need to pass the test immediately. The consequences is immediate. Compare it to design consequences, it seems like an eternity, especially if the design changes never have any test/measure to get passed.
Designer can blame developers when dev cant code their design which is so delightful the sales will improve. But developers cannot blame designer when the sales does not come through from the changes. Sales manager do.
There is test for design and it was called 'split testing'. Unfortunately it is marketers/sales job to come up with the test, not designers.
By these rules, designers should be in sales team instead of product team!