Disclaimer: I'm terrible at organic chemistry.
As with software engineering, you develop a general sense (I didn't) of what might work and what probably won't. You aren't coming at it blind and reinventing the wheel every time. You learn to recognize patterns in chemical structures and reason about how they will interact under various conditions based on that. You do electron pushing in your head without giving it much thought, similar to a programmer reasoning about object lifetimes or dataflow in an application.
As to -100 C or +5 atm, that's the easy part. You alter environmental conditions when what you're working with is too reactive, or not reactive enough, or you have some other general problem. It's roughly analogous to determining the minimum amount of RAM the machine hosting your production database requires.
That part, at least in theory, is easy to understand.
Chemical reactions are typically of the form:
ingredients + energy -> products + byproducts
If the energy is on the left-hand side, making the environment warm makes the reaction go faster. If energy comes out (so it's on the right-hand side), making it cooler is better. You also need to control the temperature to be in a range where both the ingredients and products can survive.
As for the pressure, if the products and by-products have more total volume than the ingredients, low pressure is good. If it's the other way round, high pressure encourages the reaction.
The more complicated part is that if multiple reactions can happen with the same ingredients, or if the products can do further, unwanted reactions. Then you have to balance out the parameters to encourage just the reaction you want, and to discourage all the others.
Back then, you had a certain number of mechanical properties in mind as well as well-known "parts" (gears, linkages, etc) that you could predict the properties of very well - and the task was then to assemble them into larger mechanisms that did what you wanted them to do.
Seems to me, the intuition here could be similar - except the number of dimensions in which parts can interact is larger, the "clockworks" are orders of magnitude more complex - and your tools are much more coarse, so mostly, even if you know what you need to build, the building itself can only be done indirectly.
I have a bachelor in chemistry and ochem was one of the most painful classes based on the amount of things you must know by heart for synthesis
But great professionalism can also be developed without as much cross-functionality between subdisciplines.
Like many things, detailed efforts unique to the subprofession can be the most essential.