DoT Tick haste is calculated upon cast finish, just like crit, mastery and spellpower.

Regarding point number 1, I believe at this point it's been pretty well supported that Fel Flame's damage is simply too low to be useful unless you are moving and have no other instant casts that need to go out. Since it only adds 6 seconds to UA, as opposed to simply hardcasting UA for it's full duration, it's not worth doing for that, and it's abysmal damage makes it not worth casting for that either. Other people might have a different opinion but from my experience it's useless except as movement filler.

Regarding point number two, I haven't tried it to have hard numbers, but I would have to say that if your testing shows them being more or less equal, then just go with the simplest way, which would be the Instant SF right from the start. My gut feeling is you probably CAN see a DPS increase by using the Felhunter at first for the buff then insta-cast succy, but the timing on doing that properly while still getting a benefit from it seems dicey. I'd like to hear if anyone else has an opinion on this though, because it's a good idea in theory.

I think you are not the only one curious about this mechanic. I've seen the same question asked on a couple of forums. I did a test on my warlock for the 12.5% threshold. My lock is only level 80, though, and my gear sucks. So, I wasn't able to test any higher threshold. (I did not test the 4.167% threshold because I wanted to do the test with constant spell power.)

With 11.99% haste, my bane of agony ticked 13 times: (1.786s interval)
4 weak
4 average
5 strong

With 12.50% haste, my bane of agony ticked 13 times: (1.778s interval)
4 weak
3 average
6 strong

With 12.93% haste, my bane of agony ticked 14 times: (1.771s interval)
4 weak
5 average
4 strong
1 very strong

weak = 312
average = 380
strong = 448
very strong = 515

So the total damage at each value of haste was:
11.99% - 5008
12.50% - 5076
12.93% - 5455

I tried the same test with a much lower spell power, to see if the calculation somehow depended on the total damage of the spell, and, it does:

With 11.99% haste, my bane of agony ticked 13 times: (1.786s interval)
4 weak
4 average
5 strong

With 12.50% haste, my bane of agony ticked 13 times: (1.778s interval)
4 weak
4 average
5 strong

With 12.93% haste, my bane of agony ticked 14 times: (1.771s interval)
4 weak
5 average
4 strong
1 very strong

weak = 252
average = 319
strong = 387
very strong = 454

So the total damage at each value of haste was:
11.99% - 4219
12.50% - 4219
12.93% - 4605

Now, this is very interesting. A "very strong" tick exists after we get to this new value of haste. Over the course of a 300 second fight, would this make a difference?

With 12.5% haste we need to cast BoA 12.98 times (23.1s duration) - you'll clip the last strong tick
With 12.93% haste we need to cast BoA 12.10 times (24.8s duration) - you'll get 1 extra weak tick

Using both trials, we can calculate the number of ticks at each value (weak, average, strong, very strong) and find total BoA damage in the fight for each:
Trial 1 -
12.50% - 65540
12.93% - 65772

Trial 2 -
12.50% - 54460
12.93% - 55512

So, the amount of extra damage you get on BoA by adding that .43% haste varies depending on how your spell power adjusts the ticks. It's a strange mechanic.

Jmickey is right, Lightweave Embroidery gives 580 Intellect.
While the tooltip reads "spell power", you can easily verify ingame it indeed gives Intellect. This is the buff you gain: Lightweave.