The essence of it is in good application of macros. Macroassemblers were essential tools for software that wanted all three of complex, high-performance, and fast to build on hardware of that era. Although it's a bit intimidating to write a large program without a higher-level syntax backing it, it's possible to invent coding conventions and reuse strategies that get you to the desired level of flexibility and maintainability, but perhaps without as many built-in checks and a simpler strategy with respect to resource usage(a general-purpose "malloc" is not a thing you would want to do casually in an assembly program since it makes everything considerably harder to track - but simpler fixed-buffer or mark-and-release strategies were fair game).

And if you were a little more ambitious you could also write a custom code generator that added whatever additional features you needed - expansions, type checks, expression evaluators, simple optimizations. Code generation is actually far easier today since the scripting languages now are so much more powerful and feature-rich than what a typical early 90's setup would have(assembly, buggy proprietary C or Pascal compiler, Basic variants).

Of course there's nothing stopping you from taking a similar approach now and custom-generating some source instead of using templates or compile-time evaluation or other fancy features - it's just harder to communicate.