I experienced what sounds like exactly the same behaviour with one control line speed model, eons ago, and more recently, one pattern R/C model. I must have built at least 30 control line models with their wings and stabs in line, with zero incidence, and only one acted up. The pattern job would suddenly nose down while flying straight and level at speed, losing about ten feet of altitude, then behave perfectly for ten minutes or longer.

What is happening, I think, is that the narrow airflow stagnation zone behind the centerline of the wing suddenly flips from the top to the bottom of the stab, or vice versa, exerting a very noticeable influence on the lift or downforce developed by the stab. I removed the gremlin in the pattern ship by simply shimming the trailing edge of the wing about 1/8 inch down, effectively adding about a half degree of positive incidence, putting the wing slightly out of line with the stab. I don't put the stab exactly behind the wing any more, and the problem has not recurred.