Just wanted to point out that parsing multiple topics is not easy without backtracking. Consider this example:
term1
term2
...
term999
: definition
In the above case a parser needs to consume 999 lines to detect a definition list. While in this case:
term1
term2
...
term999
next paragraph
the parser consumed 999 lines and detected that there is no definition list and it needs to backtrack and reparse the whole block element as something else (possibly a plain text paragraph).
AFAIK Pandoc also requires that a definition line is indented four spaces. Such as:
term1
:···definition
because otherwise it is all too easy to make an unintended definition list.