I'm going along with the "all results since the sample was taken" majority but I would also be OK with "take the lot away". All samples should remain legitimately subject to retrospective testing for any substance or method banned at the time the sample was taken. Cheating is cheating.
As far as the problem that creates by promoting other riders who may equally have doped, one option would be to;
* leave the positions as they were, just scrap the positions of the dopers.
* promote the riders
if they request it, on the condition that any and all samples they have
ever provided are re-tested at their cost.
Using this scenario for 2006 TDF, the two possible outcomes I can see are (A) Oscar came second (thanks very much but don't retest my samples)
or (B) the top ten looks like this:
1 ****rider sanctioned for doping violation****
2 ****rider sanctioned for doping violation****
3 ****rider sanctioned for doping violation****
4 ****rider sanctioned for doping violation****
5 ****rider sanctioned for doping violation****
6 ****rider sanctioned for doping violation****
7 ****rider sanctioned for doping violation****
8 ****rider sanctioned for doping violation****
9 ****rider sanctioned for doping violation****
10 ****rider sanctioned for doping violation****
Repayment of winnings and payment of fines is a tricky one - did Vino ever pay his fine for breaking the ProTour doping code (i.e. pay back salary)? How would this work for broke bums like Landis? What about winnings shared with the team? I could see civil action by riders to recover winnings from dopers becoming a real mess, except that few would want such scrutiny on their own results as this would bring.
What I do think should happen for any future cases like Valverde is that any rider who is riding pending a sanction should have all winnings held in escrow until the sanction is dismissed or upheld, at which time they can be paid to the rightful winner (subject to that person passing controls as above).