It's so crazy it's beyond even poetic. Firstly now ex-Senator Bob Day takes the government to court using money that may not even exist to argue that the new Senate system has prevented his re-election. Then he proves his own case multiply wrong by winning. Then it turns out he might not be eligible to keep his seat and he says he'll resign, then he flags that he'll hang around a bit. Then he resigns, and then it turns out that he might never have won in the first place - for a different reason to the one he first flagged resigning over. And it has been cited as a factor in the blowup between George Brandis and the former Solicitor-General, Justin Gleeson (though it's now emerging that that was erroneous and advice involving a different Senator, Rod Culleton, was the issue there.)
What happens now? Firstly, while Day's seat remains vacant until we find out whether it is a recount or a casual vacancy appointment that will fill it, the Coalition benefits. The Senate is reduced from 76 seats to 75, meaning that a majority is now 38 not 39, which is effectively the same as having Bob Day automatically voting with them on everything, with the added bonus of him not even being there to do it. They're probably hoping the court has some really long adjournments.
(Update: The paragraph above was written before Culleton threw a spanner in the works by flagging his intention to abstain on contentious legislation while his own eligibility is sorted. If Culleton abstains then the combined absence of Day and Culleton is very harmful to the Coalition, meaning they need 8/9 non-Green crossbench votes instead of 9/11.)
Secondly, if it is found Day was validly elected in the first place, then the SA Parliament will nominate his replacement, who by almost universally followed convention will be whoever Family First nominates. But if Day wasn't eligible in the first place, then it would seem we are following the same road as in the Robert Wood and Heather Hill cases - to a full recount of the SA ballot, minus Bob. That might sound arduous, but all the ballots are computer-entered and it's very likely the computer counting system will have provision for excluding a candidate right off the top without even needing any recoding. So it's just another button press event.
Antony Green has covered the issue in another piece that is essential reading. I will say however that I do not believe the High Court would go near the argument that Day's ineligibility (if he was ineligible) should cause Family First above the line votes to pass directly to the next party. The High Court stated clearly in the Wood case (see also here) that a disqualified candidate is to be treated the same way as a deceased one - S 273 (27) applies and the vote passes to the next available candidate.
It's true that in the new Senate system voters choose which party they will put second and third and so on (if voting above the line) rather than having that determined by the group voting ticket submitted by their party. But if having a lead candidate ruled ineligible was reason enough to rule a party ineligible to receive ticket votes, then the same ruling could have been made under the old system (in which case the votes would have passed to whatever party was second on the disqualified party's GVT). So I can't see any material difference with the Wood and Hill cases.
I'm expecting very strongly that if there is a recount, Family First #1 above-the-line votes will be allowed to flow to the Family First #2 candidate, Lisa Gichuhi, as will Bob Day's #1s that are #2 for her.
In a recount of this kind, the whole Senate election for the state is recounted. It is not the same as a Tasmanian House of Assembly casual-vacancy recount at which only the vacating candidate's votes are recounted.
The major question would be of the final seat between Gichuhi and Labor's Anne McEwen. This is only a question at all because Day's margin of victory over McEwen was relatively small (3543 votes). Day himself had 5495 below-the-line votes and would have received several thousand more in preferences, though I don't have the exact number. (They include 1233 that were #1 votes for candidates who were elected very early; these would seem unlikely to flow much to Gichuhi.) If a very substantial percentage of these turned out to be personal votes and preferences for Day that did not flow to Gichuhi, then that might cause Day to be replaced by McEwen, which would be a complete disaster for the Government.
On the other hand, Gichuhi gets back a small number of votes that in the original election were 1 for her but didn't flow to Day.
Grahame Bowland has simulated the issue. In his simulation (which in this case should be more or less exactly what would happen as there are no serious interpretation issues*) Gichuhi does still beat McEwen for the final seat, by 3601 votes.
On the surface, this makes no sense - why should Gichuhi win by more? It turns out there's a reason for it. In the original count, after the exclusion of Sean Edwards (Liberal), Bob Day led McEwen by 1439 votes. In the recount at this point, Gichuhi trails McEwen by 1372. That 2811-vote turnaround (caused by Gichuhi getting fewer primaries and preferences than Day) would seem to be almost enough to wipe out Day's original margin, cutting it to 732 votes with later transfers perhaps bringing it down further.
But here's the catch. In the original count, Senators Hanson-Young and Kakoschke-Moore are 16 and 454 votes short of election at the point where Edwards is excluded. They are elected on the votes of David Burgess, with substantial surpluses which advantage McEwen compared to Day by 4272 votes.
But in the simulated recount, the release of Bob Day preferences that don't flow to Gichuhi puts Hanson-Young and Kakoschke-Moore over quota once Edwards is excluded (by 466 and 35 votes respectively), meaning that their surpluses are actually very small and only advantage McEwen over Gichuhi by 245 votes. A surprising difference, but one that places Gichuhi beyond risk.
So we can put some of the speculation away - unless the High Court does something really strange, a recount would just see one Family First Senator replacing another.
(Note: If they do rule the whole FF ticket ineligible, then Bowland's simulations have shown the recount electing McEwen (ALP). The scenario in which Gichuhi is also ineligible has not been simulated but it looks like McEwen wins there too. No scenario has been found in which the Liberals or One Nation win the seat.)
* Obscure technical note added Nov 3: There is some potential ambiguity about whether below-the-line votes that were originally formal but contained some duplicate or omitted numbers might become "more formal" in the recount. For instance if a below-the-line voter numbered all the boxes but put both Day and Gichuhi tenth, this vote would originally have not reached either, but in the recount might flow to Gichuhi. The recent release of the AEC's preference distributions for the advisory Section 282 recounts (see earlier article Majors Stitch Up Senate Term Lengths, Film At 11) has created some surprise because they seem to have allowed votes to become "more formal" in such ways. However, the number of such votes involved is likely to be trivial. Even if, for instance, there were enough to cause Kakoschke-Moore to not reach quota on the Edwards exclusion, Gichuhi would still win.