I wouldn't even call this bullying yet, at this age, just a child with a lack of social skills acting out
Although some might say that acting out would tend to occur in interactions with various children, therefore form a pattern which in general may be noticed by the teacher. From the OP, it appears the perp's actions may be directed at one child, who may be targeted for the aggression. Additionally the perp may choose his timing so that his aggressions remain covert, characteristic of deceit/manipulation.
And, though not coming down one way or the other on this specific case, I will point out that antisocial behavior patterns have as much predictive accuracy as intelligence at age 3. (In a fifteen minute playground observation.) That is, the relational profile is as stable from age 3 to 16 as measures of intelligence are. (Not very stable, but somewhat.) Prosocial/antisocial skills become stable at about the same age as intelligence, too (around age 8-10), meaning that antisocial behavior patterns become entrenched and largely intractable to treatment around the end of third grade. (See the work of Hill Walker.)
Thus the lack of teacher intervention in the behavior/lack of social skills of the aggressor is extremely significant in the developmental trajectory of that child, and their likely future isolation from prosocial peers.
