Regarding Dieter's second question, there are the "Nebencaesuren" of āryā-type meters, which are not (to my knowledge) described as yati by the tradition: in certain positions (the sixth gaṇa of the first line, and the fourth gaṇa of any vipulā line, as H. Jacobi discovered), there must be a "syncopated" rhythm, which is achieved either with the shape
v–v or with v|vvvv, the word-boundary being necessary in the second case to prevent the four light syllables from being scanned as two bimoraic feet. Correspondingly there are "bridges" where, in order to ensure that the last and last-but-one mātrās of a gaṇa are scanned as a single bimoraic foot, no word-boundary is permitted between them. You can read more in C. Mayrhofer in IIJ 31 (1988) 17-25 and my paper on the āryā here.