Update, May 30 2013:
Several issues have been brought out by the situation in the previous post regarding filter 5, which is set up to filter roofwater. Roofwater can vary considerably depending on the physical, and geographical location. Temperature, acid rain, debris, animal and bird droppings, pollen, and dust will greatly influence the water quality.
1. If the input water will periodically increase in turbidity, the sand size must not be smaller than .25 mm effective size. If the input water is consistently very non-turbid, being less than 2 or 3 NTU, then .15 mm effective size may be ok. A value of .35 mm effective size for all the sand in the filter is the best overall choice if the input water is subject to drastic changes in turbidity at times. Fine grained sand will clog up very quickly if the input water becomes more turbid (cloudy – containing microscopic particles of dust, pollen, or bits of debris from roofing material), and this creates the need to clean the top sand layer often, resulting in very poor performance and increased opportunity for anaerobic conditions to start up in the filter.
2. Most importantly (this is extremely important), re-circulation of filtered water through the slow sand filter is absolutely necessary, if rainwater/roofwater is to be filtered. These filters cannot be allowed to sit idle with no flow. The filters here that malfunctioned and became anaerobic did not have automatic re-circulation of filtered water; and I neglected to add filtered water daily. The water I was adding to filter 5 was from an open bucket which had become contaminated, and I was not adding water often enough. Furthermore, the water I was adding did compound the problem because of the high turbidity level of input water due to the diverter malfunctioning combined with the high turbidity level of the contaminated water from the open bucket. These filters are not foolproof.