I just tested a S9 I have running on solo right now with a backup pool of Kano. In console I switched to kano and timed the difficulty adjust. Diff stayed at 1024 for exactly 60 seconds and then switched to 8.38 K. Before I had done this test it was running at 10.8K on solo and had been for days. When I switched back to solo diff quickly dropped down from 10.8K to 4K after only a few minutes of no hashrate. I understand about the smaller miners but as you said they will probably never make money mining, so why cater to these in-efficient miners?
Not sure how you're conducting your tests for you see... solo and kano run the same pool software which I wrote so what you say is nigh on impossible. The only difference is solo's start is exactly 1k whereas it's 1042 on kano which is virtually the same (+ Douglas Adams' 42)
I understand you wrote the software that is why I am making the suggestion. It is not hard to ssh into a S7 or S9 and switch pools manually and watch the share submission and time difficulty adjustment. I can take a video of it if you want I guess...