1. If you have already converted the Poser characters and their morphs into LW objects with endomorphs, then you should be fine. You can also load collections of .OBJ files (one base and one per morph shape), if you need to work with models that haven't been LW converted, but since you'll be animating in LW, that shouldn't come up. (TAFA includes some tools for fixing orientation issues when working with objects from different toolchains, should it be necessary)
2. That will depend on how hot your machine is (both CPU and GPU). TAFA was originally designed for medium-poly control cages that are subdivided, and it does that subdividing in realtime as it animates. Since you will be using much higher polygon models, you will want to set the subdivision level to 0 so that it doesn't try to increase the polygon count any further.
3. First, "splitting the morphs" is only an issue if you wish to apply a morph asymmetrically. Say that you have a single morph that has both eyes closed. You might want to asymmetrically apply just the left side of it to provide a wink. If you are willing to use MDD files (which are much larger, but let TAFA do splitting without modifying your morph shapes), the MDD files specify the location of every vertex for every frame, so TAFA can specify any asymmetric balance you need. If you need to use Morph Mixer files (which are much smaller), then TAFA can only specify a single weight for each existing morph -- which would require you to split your existing morphs into left/right pairs before TAFA could apply them "asymmetrically" (it isn't really asymmetrically -- TAFA has to apply the new morphs fully, but since the morphs themselves are asymmetric, the end result is the same).
3A. As a side note, TAFA's morph splitting was also designed with the idea of control cages that get subdivided. The subdivision normally helps smooth out some of the transition zone between the left and right half of the model. With a high polygon model and no subdivision, some morphs may not be as useful for asymmetric application.
4. No. We've found that artists seeking cinema quality animation can actually produce the full animation data faster with TAFA than they can clean up and tweak the phonemes placed by automatic analyzers. This result is dependent on how picky you are about the look of your character, however

If you are working on a daily 30 minute show, it may be more important to get the lipsync slapped in very very fast, even if it isn't as good -- in which case, unfortunately, TAFA isn't really the right tool. But if you care more about your characters looking right, TAFA provides several useful tools. One very useful tool is the ability to specify a punch-in/out region that it will play in a continuous loop. You can drop in new morphs, move morphs, tweak morph strengths and balances, and it just keeps playing that loop showing you what things look like now. It makes tweaking your animation very fast and easy. TAFA also doesn't require any initial mapping from analyzer morph names to your morph names (but it does provide for changing morphs (or characters!) in a target after a scene has been set up -- you can remap old morphs to new morphs and have it applied to the whole scene, which can save an immense amount of time. Lip syncing can start before final models are ready without losing work or requiring much rework, because the scene done with the early character can simply be remapped to the full up character. It's a life saver for those late-in-the-project changes that would otherwise kill you.)
5. The current version is v1.2.2.8. TAFA checks for updates itself, and you will receive any updates that are short of a major rewrite. As you look around on the board, you'll see that a major rewrite is not likely to pop up anytime soon (well, except that I'm now working on an XSI plugin version, but that won't affect you). Certainly bug fixes and the like will be free.
6. I have a long list of features I'd like to work on, but I honestly don't place any of them (other than the XSI plugin) in the "near future" category, simply because TAFA is a home business and isn't what pays for my bills -- which necessarily puts it at a lower priority.
(Demo) While I understand the aversion to demos, I encourage you to try the TAFA demo for a few reasons:
A. You'll need it anyway if you do an online purchase -- the online purchase merely turns the demo into the full version.
B. It's the absolute best way to know if you'll have any problems with your models and performance, because the only missing feature is the ability to save files (which does mean that you won't want to spend a lot of time with the intent of serious output, but you could spend some time piddling around to see if it is fast/smooth/easy/intuitive enough).
C. I would recommend trying things in this order:
i. Load your model and see if basic loading and model rotation seems smooth enough. Don't forget to switch to 0-subdivision with the buttons along the top.
ii. Go through at least the basic tutorial with the supplied models (or use your own -- the only difference will be that the morphs in your model won't match the morphs mentioned in the tutorials, but you can work around that). The tutorial should only take 1-2 hours.
iii. If you did the tutorials with the supplied models, go back a play a bit with a short sentence with your own model to see if animation speed is acceptable.