I've worked with R&D projects (non-software) and there are similar patterns. Early stages involved various proofs of concept to eliminate research with a low chance of success. The goal was to "fail early" and preserve resources for projects with the highest probability of success. Instead of a spiral, we had a linear path with critical stage gates at which each project would either fail or pass. Each successful stage-gate opened the project to more funding until we moved to the final push to full production.
Projects had to compete for resources, so our portfolio was shaped like a funnel: large at the early stages with a maximum number of new ideas, and smaller as successful projects getting closer to completion.