I was reading Why can't "fi" be separated when being copied from a compiled pdf? and had a thought:
I know that v1.4 and up PDF documents have an OCR layer. Would it be possible to have PDFTeX or luaTeX place the un-ligatured text into the OCR layer, so that you don't have inconstant or odd behaviour when copying ligatures?
It also occurs to me that this also be a workaround for my troubles copying mathmode greek characters, discussed in Proper way to use greek letters in an English document . This also seemed like an ideal way to provide more accessible code for equations, as trying to copy them right now is not exactly useful.