Most of the presumed problems this proposal intends to resolve originate from the assumption that semantic TEI elements like "p" (paragraph) are (also) defined by layout.
That is why the code examples do not use "lb" at the beginning of such elements - there is the underlying assumption that a "p" implies a line beginning at its own start, and similarly all the other mentioned elements like "dateline", "salute", "signed", "opener", "closer", "address", "addrLine".
I strongly disagree with this view. In such an encoding, it becomes difficult to determine the physical lines. If a documentation of physical lines is desired (which is not true for many editorial projects), it should - in my opinion - be done always explicitly, even at the beginning of elements like "p", by inserting the "lb" element ("line beginning").
Similarly, the encoding of a verse by the "l" element (indeed very misleadingly labeled "verse line") does in fact not imply that the verse starts on a new line (nor that it ends at a physical line end).
So some of the proposed solutions in the following text - such as "@rend="inline" proposed for "salute", "signed" and "dateline" where the text does not start on a new line - are really completely unnecessary if the "lb" element were used consistently for all line beginnings.
While many "chunk" elements may often coincide with layout features like a new zone, a new line or an indentation, they are independent of these features, they are not defined by them. A new paragraph (conceived semantically) can very well start in the middle of a line, and it is really up to the editor to determine where a textual section like a paragraph starts (or ends).
It is my firm conviction that a clear conceptual separation of semantic elements (and most of the TEI elements are semantic elements) from the documentation of physical layout would clarify many problems which are dealt with in this proposal.