The following process seems to be a recurring one:
Initial design. Commercial organisations invent and implement a design that seems at the time to work.
Problem discovered, new design suggested. Due to new developments, or just because the designers missed something, a problem with the design is discovered. Academics find a solution to the problem, and a new design is proposed.
Old, sub-optimal design persists due to existing production. However, the original sub-optimal design is maintained, because a large amount of devices with the old design are already deployed, and for various reasons (network effects, the need for compatibility between devices, etc), this means producers don't want to switch. Or alternatively, new production switches, but we still have to deal with the old design because of the large already produced devices.
I am looking for historical examples of such a process in the context of the IT security of CPU/instruction-set-architecture:
What designs of instruction sets/CPU's are we stuck with that are bad from a security perspective, but that are maintained because the cost of switching is high? (I can imagine that this is also intertwined with OS design and internet protocols).