The main difference between a TEXT solution and a precomputed solution is that the TEXT solution must have used exact knowledge about the contents of the judge data.
The submitter of TEXT knows exactly what the judge output lines are and their relative order. The precomputed solution does not rely on knowledge of the judge data, since the user is reading the input and producing corresponding output. It also uses a list of all possible cases, not just the ones in the output file. The user may have used a slow/suboptimal solution to come up with the precomputed values, but it is still the user's own solution (unless you have proof otherwise) and can not be called cheating.
Frankly, I don't understand why TEXT is selected by default in the list of allowed languages in the problem editing page. Except for a few rare and unconventional problems problems (e.g.problems with no input like BCEASY), I see no way that someone could submit a TEXT solution and not know anything about the judge output.
DOC, PDF and PS are deselected by default, maybe deselecting TEXT is also a good idea to prevent cheating.