Sounds promising. If the opto coupler's current consumption is unwanted at the high voltage node, you may want to add an
auxilary low voltage supply to power it. But 1 or 2 mA should be sufficient.
when you place your resistors as part of your Vsense, calculate how much voltage drop is across each resistor. you will find that you may need 6 resistors in series.
i usually don't use optocouplers. instead i just go to a couple of op-amps in differential measure mode and that's good enough. used succesfullly up to 600V on the bus in high noise environments.
keep in mind.. what happens if you lose a resistor. will the voltage still be attenuated? or will you suddenly have high voltage on the input pin of your optocoupler? you need a way of beign able to lose any ONE resistor in the attentuation path which will cause a FAULT trip OFF event.. without putting super high voltage on the pin of sensative component (op-amp, optocoupler etc). think about it..
product reliability is what separates a hobby project from a professional product.
Usually, resistors are assumed reliable in contrast to semiconductor devices.
A suitable assembly technology and sufficient power rating is required, however.
If you need insulation for safety (especially reinforced), I would still use an insulator. Phototransistor optocoupler with TL431 would be a conveninet solution. There are many solution notes about that.
If you're referring to safety risks, the feedback loop surely can't achieve a safe limitation of converter output voltage, you need independent means anyway. In so far, possible damage of parts by a disconnected resistor isn't particularly a safety risk. The important point with resistors is, that you don't need to fear a short circuit fault, as long as you keep the rating. In case of a fusing resistor, you can even expect a safe behaviour with overload.