I wrote both test bench and aircraft code to replace an F/A-18 (C/D) chip in the RPYC (Roll Pitch Yaw Computer) due to the original being NLA. If I recall correctly (this was a couple decades ago now) the new chip was NEARLY a drop in but had a slightly different rise/fall curve and a couple of ns faster response. It was the closest I could find but required software updates to both the aircraft and the test bench to not be an issue.
So yeah, this has been an issue for decades. I'm personally glad to see a company addressing that market.
Talking of mainframes, many core bank registry functions are emulations of prior systems long embedded into architectures now themselves superseded. Support for tech archaisms has long roots.
I'm slightly surprised that _virtually_ identical chips are allowed. I expected truly identical chips to be required…