This is not thought out as well as the other two examples.

Consider this; a railway (railroad) system or a highway system or an airport or ... well you get the idea, are basically static.  Things move around them but the infrastructure remains constant.  Parts of the infrastructure can get damaged or even fail completely.  Also the likelihood of any part of the infrastructure failing within a set time period can usually be calculated.  If each item in the infrastructure mapped to a cell in the infinite maze then pseudo random failures could be simulated according to their likely probabilities. 

The benefits of using the infinite maze to generate these failures rather then the usual methods are that the maze is big enough to map each railway sleeper or each road surface reflector or each landing light.  Assuming this was done in a two dimensiontial plane then the third dimension could be scanned to generated the errors.  The third dimension is large enough for the time period to be microsecond or even smaller. You could use it to simulate electronic circuits.

Then if you wanted to re-examine a particular set of failures then you can start from the same place in the third dimension and all the errors will re-occur in the same sequence.