Rationale (June, 2014):

CWG was sympathetic to this request, although it felt that the conversion to a function pointer was irrelevant to this question. However, EWG should examine this request for a language change before any action is taken.

[Accepted at the November, 2017 meeting as part of paper P0624R2.]

It would be more consistent if the closure type of a lambda expression with no lambda-capture were default-constructible and copy-assignable, since such lambdas are already convertible to a function pointer, which has those characteristics.