As a software engineer, I wonder how accurate they stuck to their said schedule at the last page.
It was planned to be a totally 2D game in the vein of Diablo, but as this brief lays out, the amount of 2D assets required for all the rotations of the characters gets out of hand extremely quickly. After we started development we figured this out and decided to switch the characters to 3D and keep the backgrounds 2D. So I had to write a whole 3D engine and toolchain from scratch in two weeks. Those were the days.
Still, took three years to get that pig out of the door.
Just insane amounts of work and data. There's a cool video they did about the part of the process that comes before that - creating the models from the character artist's 2D art. https://www.youtube.com/watch?v=cYJ6d1ifSqA
Switching to 3D gave us unlimited movement but bought us multiple other problems like low poly count, lack of solid integration with 2D backgrounds, and difficulties in combining huge 2D and 3D rendering pipelines.
It's more like lean product development of a car or other product: you have design targets and may do concurrent engineering / spikes to explore the design space iteratively, but eventually you converge on what will be achievable (or you get stuck in game development hell), which point its a matter of bug fix triage and endless tweaks from game testing feedback. It's iterative but not what one would call "agile"