The only failuress of 485 networks I have seen has been at installation and in the lab due to faulty wiring (voltage on the 485 line and ground potential differences between the ends). I have 485 networks in heavy industry locations and none have failed. The fact is that the likelyhood of failure of a well designed galvanically isolated 485 network is extremely small. Maybe my success is due to the fact that I use isolated 485 transcievers such as the MAX1480 which can handle 1260V between the signal line and ground (for 1 minute, I just looked at a typical datasheet). Prior to the use/availability of isolated 485 transcievers we had some failures in the lab and at installation due to greatly different ground potentials between the two ends.
I know of no way to make a 485 network automatically disconnect a failed node and when the precautions above (isolated transsceiver with the ability to handle crazy ground potential differences, ESD protection, thermal shutdown transmitter) are taken the need for designing automatic disconnect is the same as the need for hurricane insurance in antarctica.

Erik