A botnet might not be suitable.
Let's run some (very rought) numbers:
Network hashrate is around 4-5GH.
Rough calculations say you need about 16 MHash/s to find a block a day on average.
That's roughly 1600 machines if you're CPU-mining (assuming 10 kHash/s per machine) or maybe 50 machines if you have decent GPUs in them. Those GPUs will consume around 10 kW, so that'll net your school district around 240 kwH per day (with European electricity prices, we're looking at roughly €50 per day (for 12.5 AUR worth about 0.5 €).
Maybe better get one of these:
4 of those cubes with a controller and stuff would set you back something like 2-3000€, but would have 20 times the hashrate (300MHash/s) of the botnet described above. So that thing would net you roughly 250 AUR/day (worth 10 €) consuming €5 in power (with the assumptions from above)
We can deduce from these (thought and practical) experiments: mining AUR is a very competitive industry.
I was mostly joking, but I could tell you that I have domain admin rights to around 600 student desktop computers which aren't used at night. Half of those are in graphics arts labs, each with dedicated video cards. I'd be in and out before they got the power bill, and even then, they wouldn't know what caused the spike. Not that it's possible, but if there was a way to mine on a Chromebook without tipping anyone off on what was going on, I could push a miner to around 10k Chromebooks.
I manage both domains, so it wouldn't be difficult. And worst case scenario, I could just claim plausible deniability or say it was some kind of virus. The upper administration here doesn't even know how to use excel. You could tell them that the reason their wireless keyboard doesn't work is because their CPU died, and they would think it's totally plausible. Gotta love the California public education system. All in all though, it's something I've debated with some coworkers before... at least when scrypt asics were just coming out. Now though, it would just be a test in theoretical possibilities.
The titan is a great machine, but even the ones being sold are being listed with bad dies. I've read a lot about the failure rate on them over the last few months as well... it's not something I would want to purchase used. That's like a ticking timebomb of profit loss. But even if you did buy it used, still, that's $3k, and then at least for me, something to the tune of $300-400 in power every month. California has a tiered rate system that puts you up into a $0.36kw/hr rate if you run, what seems like, more than a fan and a lightbulb in your house. Even with the bitlicense issues in New York, I'm going to be happy to move back there in 2017. Flat rate $0.11-0.12 where I'll be, so I can turn my 20mh in scrypt gear back on.
has a crypto currency ever succesfully swithced from scrypt to sha 256?
Not in this day in age, I don't think. You would have to hard code a starting difficulty to mitigate the initial hashrate increase. Even then, you're playing a guessing game on how many hashes would swing over and possibly stall the chain. There are better options out there though, rather than just going straight to SHA. Rocanonz mentioned one of them. But like I said... we discussed almost all of our available options, and we're working towards a few of them right now.
-Fuse
.
long time lurker, but eager to see this prosper. I look forward to being part of this marathon, and believe this can become what it was meant to be, given enough energy and time. 
