That very much depends on a host of things. You can often get away with brewing at high cloramine leveles for quite some time and then all of a sudden you get hit. Remember that it is a chorinated phenolic that ruins the beer not chloramine and there are lots and lots of phenols available depending on what may be in the water, what's in the grain, how effectively the phenols are extracted from the grain, how much chloramine may have escaped from the water and whether conditions for reaction of chlorine and phenol are 'favorable' for chlorphenolic formation. It's probably safe to say that if you grind your grain (and hence husks) to powder and sparge with very hot water at high pH your chances of chlorphenolics with highly chloraminated water are better than if you used less chlroaminated water and treated the grains more gently but I wouldn't want to go beyond that. As chloramine is so easily dealt with it just makes sense to treat for it if you have it rather than try to predict whether you will get away without treating for it.
As to detection thresholds: note that Martin said 'detectable chlorophenol level in beer is on the order of 10 ppb'. This is not a very precise specification nor is it intended to be. Values of 1 to 100 ppb are within 1 order of magnitude (and thus on the order of) 10 ppb because their logs differ from log(10) by 1. Thus detectability is quite variable depending on the particular chlorphenolic, the matrix (beer) and the taster.
How do you get taste thresholds? By having panels taste. What they taste depends on what you are looking for. To determine a taste threshold for salt, for example, you could prepare a solution of a given concentration and present it to a panel to see how many detect the salt. This is often done by presenting triplets of cups to each panel member. At least one, but not more than 2, of the three cups will contain the salted water and the other(s) plain water. The panelists are asked to declare which of their three cups is different from the other 2 and to tell whether the different one tastes salty or not. The results are scored and the results tablulated in the form 'x% of panelists were able to detect y mg/L NaCl at the z% confidence level' with the confidence level being the probability that x% could have been arrived at by coin flipping. The investigators have flexibility in what they want to call 'detectable'. If 5 out of 20 panelists identify the different cup correctly then you probably wouldn't call y mg/L detectable because the probability that they were coin tossing (panelists are told they must pick one of the three cups as different) is 85%. If, OTOH, 14 picked the different cup correctly, you probably would call y mg/L detectable because the probability that 14 out of 20 coin tossers would get it right is 0.08%.
If you got 5 out of 20 the obvious next step would be to repeat the test with 1.5*y or 2*y or some other multiple of the original concentration. Suppose you pick 2*y mg/L and now get 14 out of 20. You are now pretty sure that 2*y is above the threshold and that y isn't. Thus the threshold must be between y and 2*y but where? You could now try 1.5*y.....
As you can see you can wind up doing a whole lot of testing and still not come up with a solid number. If you consider the problem of trying to determine the MLD for a drug or poison you will appreciate that investigators have evolved techniques for minimizing the number of tests required but even so a hard and fast number is difficult to come up with. That's why it's 'order of 10 ppb' or 'at ppb levels'.
