1. Both the NTK and the Bosch run optimally at 750-800 deg C Sensor Temp. If the sensor head gets heated by the exhaust to that temperature there is no more margin for the heater controller and it runs essentially 'open loop', shutting the heater off. That's why a cooler location is better. Specially because the Bosch Sensor is more sensitive to temperature than the NTK (but less sensitive to pressure) it would result in large errors when running without tight temperature control.
2. (And more important) The Bosch sensor has a parasitic resistance in the common path between the pump-cell and the measurement cell. This resistance has a positive temperature coefficient (resistance increases with temp) and is dependent on the sensor's housing temperature. Because it is buried inside the sensor it cannot be directly measured. When measuring the sense-cell impedance or pump-cell impedance it is essentially in series with either. Both, sense-cell and pump-cell impedance, have a negative temperature coefficient. Because the heater control of the Bosch unit depends on either the sense-cell or pump-cell impedance, this parasitic resistance can disrupt the temperature control by essentially compensating the negative coeff. of either when the housing gets too hot. This means that no temperature control is possible and the resultant values would be all over the map. The LM-1 can detect that condition and throws a sensor timing error because the sensor timing with our circuit is one of the variables that is dependent on the sensor housing temp. That's why keeping the sensor housing as cool as possible is important.
Hope this explains the reasoning.
As for concerns about response time further down the pipe:
Assume a 3 liter engine at 2000 RPM and a VE at full throttle of 1.0, an exhaust pipe of 3 inches diameter, sensor placed 6 ft from the engine, and an average exhaust temp. of 800 deg C (averaged over the entire pipe length). You can calculate that the time it takes for exhaust gas to move from the engine to the sensor is about 50 milliseconds. This is a plenty fast response. Thinner pipes, higher rpms or VE of > 1.0 (boost) of course would decrease the time. Partial or closed throttle would increase the time.