Thanks to everybody who've already voted! And for comments especially -
it's curious too see how different the motivations are (from mine, in

*Votes so far*

For dialects:

Positive - B (Dan S.), D 3x (Dan S., Elliot, Yoon Ha), E 4x (Elliot,
Taliesin, Yoon Ha, Daniel A.)

Negative - B (Daniel A.), C (Daniel A.)

For words:

B(1a) B(1c); D(1a) D(1b) D (1f) D(1g) D(1h) D(1l); E(1a)x3 E (1b) E(1d)x2
E(1g)x2 E(1h)

B(2d) D(2c)x2; E(2a) E(2b) E(2d)x2

B(1000a) D(1000a) D(1000b)x2 E(1000a)x3 E(1000d) E(1000e) E(1000f)

So, E is leading, with _vauhita_ 'one', _lief'u_ 'two', _eulfu_ 'one

Next candidate is D with a lot of choices for 'one', _leufte_ 'two',
_haliulu_ 'one thousand'.

But stats aren't too convincing yet :)

Any more votes? Please!