Looks like I wasn't clear enough. There are rumours that K32's are already broken because an institution is already able to make a plot with the requirements of a challenge. That means you can do a plot fast enough so that every challenge answer from you is to be accepted and you receive XCH. I would call this 'plotting on the fly' or just jumping the queue. This also would mean, that someone figured the Hash algo ... Isn't it so?
But they are rumors.
I understood that the lab was able to create a plot in some 30 seconds, but that this is not a safe way to win (yet), as in order to provide the "best numbers" you need to know which ones they gonna be; but even if you try to recreate the plot with the picked hash as input the result not automatically becomes the "best" plot...as the small variances generate many similar but only one (or few) best plots.
So in order to generate the best plot you'd actually need to predict which path the algorithm is going to take through a file (potentially possible dropping ML on the algo) or generate so many plots on the fly that the likelyhood of you haveing a better plot then the other ppl out there is high enough to guarantee lots of wins...
Both has not been done yet from what I gather.
For me, it looks like K32 has twice the chance to be selected as K32 takes more than half the time to be plotted, but is only half the size of K33. So the faster you plot the earlier you farm the same 'Net Space', the higher the chances to win a challenge. Or am I wrong here? the whole 'theory' got busted while expectations about how fast K32's can be plotted were underestimated. Lots of guys already go for K33 and K34 while expecting a change in protocol due to the 'f....up' under K32's ... Consider it done.
True, if you take the difference in plot times as a significant aspect (hours or days) then k32 is better since you can start farming earlier.
Otoh, given the miniscule chance of a single plot to be the winning one, I would say that the likelyhood that the newly created two k32 plots (before the k33 finishes) are the winning ones seems low enough to not have a huge impact.
At this point I'd not say there is a need to switch to k33+ - if the algo gets broken, then its moot if you run k32 or k40 - as soon as the best plot is predictable you can forget the coin.
The need to plot many potentially winning plots on the other hand is a threat, but you'd need a lot of plots to increase your chances in a meaningfull matter - given that 1/512th of the netspace is a potential winning plot at all times thats some 500k blocks competing with your's for the win...
Edit - I think the point when they or any plotter (should) switch is when (if) there is a replot for pools.
Anything new then should probably play it safe and up the ante...