WIth all that combined, it reallly is good question whether whole thing should be rethought and started from scratch...
The problem with starting from scratch is that it is very easy to 'through the baby out with the bathwater' (so to speak)
Also starting from scratch and you will get somepeople saying why bother with anything related to the QL at all, its past, its gone, its not coming back - ever
I think that none of this is relevant and missing the point entirely.
IMHO - it is the concept and ideas of SuperBasic (easily understandable, powerful yet flexible and simple enough to use) that made the QL what the QL became. SuperBASIC sat on an operating system that allowed it appropriately or otherwise access to the hardware of the QL. Users of other languages including low level machine code will have slightly different views obviously and that is perfectly acceptable, however there is a difference between what is necessary and expected (i.e keyboard, mouse, screen, memory space internal as well as external) and what is quite specific to the QL which will have to be different (i.e. other i/o, serial ports, ROM ports, RS 232, Printer etc.).
Starting from scratch conceptually might be a good thing, but unless you are going to take your 'fan base' with you, you will probably find yourself isolated. Even those who are watching how the Amiga based home brews would probably agree and see it is a very slow process. Amiga One seems to have come to a complete stop, without community support it is unlikely to get very far, gone are the days of "lets put the whole show on in the barn" enthusiasm (Mickey Rooney fans will know this) and that is all you would need.
Without cold hard cash and lots of it, starting from scratch will be slow with no quick fix. The hardware in which only a SuperBASIC style interpreted language sits on, using different processors and different low level language I have already queried in a previous thread (see here
http://www.qlforum.co.uk/viewtopic.php?f=19&t=244). Unless I have got hold of the wrong end of the stick, I think what Dave is exploring is the possible avenues that exist, using current hardware compatible (to some extent) with the previous incantations of the QL, hence compatable with most of the intricacies of low level languages + high level languanges but not with the specific hardware that really is no longer required.