The core of Krashen's Monitor Model is that subconscious "acquisition" is the primary means by which proficiency in a second language develops and that the key to this acquisition is "comprehensible input". The problem of determining exactly what constitutes comprehensible input for a given learner, then, is central to the task of second language teaching. Informally, input may be considered comprehensible if the learner seems to "get the gist" of what is being said in the target language. This description is inherently subjective, however, and how exactly this might be quantified in more precise terms is up in the air.
It seems to me that, over the course of a semester, several short diagnostic exams should be administered to students in second language classrooms. These exams should focus solely on comprehension and be timed such that very little deliberate analysis of the input can be performed. They should not affect student's grades, except possibly as extra credit. If such exams are very carefully constructed, then student's correct answers to content questions can be used as indicators of their acquisition of particular grammar points. Students' scores on these diagnostic exams would be used to determine when input containing more difficult grammar points should be introduced to the class. Of course, not all of input the students receive in the classroom will be fully comprehensible, but consistent use of this method could increase overall comprehensibility, especially in written materials.
While I am inclined to agree with Krashen's general approach, his assertion that "conscious learning makes only a small contribution to communicative ability" is a bit too strong for me. He insists that "learned" information can never be directly incorporated into the "acquired" system and is really only useful for monitoring L2 output. However, just as in first language acquisition, I believe that a learner's output feeds back into their "language acquisition device" as input. If this is true, then optimal use of the Monitor - or even heavy reliance on learned routines and patterns - should be expected to profoundly affect the acquisition process. In short, practice makes perfect.
For Krashen, the language teacher's role is not to present explicit grammatical rules or even to correct students' errors. These things are simply "not relevant to language acquisition". Instead, language teachers should be more concerned with simplifying their presentations (both written and spoken) so that everyone in class can more-or-less understand. As the natural acquisition process progresses, less and less simplification should be necessary. This method may sound a bit loose, but a structured syllabus can (and should) still be followed. Providing naturalistic input at just the right level of difficulty (both grammatical and lexical) is no mean task! I think the most important point is that organizing a syllabus and actually teaching a language class are two very different things. As teachers, we must not let the structures that we're trying to help learners acquire get in the way of that acquisition. We should be more guides than teachers in the traditional sense. As Krashen puts it, "We teach language best when we use it for what it was designed for: communication."