I don't know how hard it is. But I can certainly say there is no business inscentive for it.
When it comes to improving performance by a few ms, or implementing a new feature, business people will always choose a new feature, unless the current performance is unbearably slow (we're talking regular 1.5s+ wait times for BE response).
And it's not even a modern problem, legacy software written 20 years ago has the same latency than most modern backends from my experience.
When I worked at Cisco we had regular 10 second responses on our main flows (bad data model) and still couldn't get buy-in to fix it.