I went out to fill up with petrol just now. I get to the local petrol station before the bike's properly warmed up (2 bars showing, generally, and the bad running starts at 3 or 4 bars) so it's always running fine at that stage. Filled up with the usual and then carried on riding - it was still playing silly bu**ers but didn't seem as bad as before. Then I disconnected the lambda sensor and it improved. Not hugely, but a fair bit - to a level where I could put up with it if a bit of tweaking of the TPS improved it further. But boy, did it backfire on the over-run!
Of course this doesn't prove that the lambda sensor is at fault, necessarily - just that the fuelling without it connected makes the bike run better than with it connected. I'm not sure whether I want to cough up for a new lambda sensor just to try - I've already done that with the coil, leads etc.
I can't help wondering at the improvement that happened (and I think has happened before) when I go from an empty tank to a full one........
Scuba-Sparky, you can check the resistance of your leads - see my post above for the resistance of the new leads I bought compared to the old ones. If yours have a similar resistance then you can probably save yourself the money.
PS I discovered today that the bars/top yolk rock from side to side so there's more to deal with. For a 27k mile, main-dealer-serviced bike it seems to have had a hard life...