I'm experimenting with easy-to-use ways for a package maintainer to modify a"live" package-- i.e. while it's loaded & in use. One difficulty is that packages with namespaces are usually sealed with 'lockEnvironment', which means you can't add/remove objects. After some effort, I've managed to bypass the sealing step, and the live-editing seems to be working pretty well.
However, there are reasons for sealing (e.g. in Luke Tierney's article in R-news 3/1). Which disasters am I really courting by not sealing the namespace?
NB 1: non-sealing only happens during development by the package maintainer, so some rough edges are tolerable.
and 2: the only alternative I can think of-- and it's less satisfactory anyway-- involves changing the namespace's parent environment using 'parent.env<-', which comes with even more dire warnings...