+1 with the peerless sls woofers. 
+1 din with crossing them low, probably at 2.5khz third order. 
the tweets are reasonably flat, 1khz to well above 20khz and they can go as high as 40khz...
couple of big name brands are using the tweets on their flagships like Mission, Sonus Faber, Polk Audio, etc. that in itself says a lot about the tweets.
When coupled with bigger SLS woofers (the 10" 830668, for example), crossing the XT25s at 2kHz may be a better idea to minimize the audibility of cone breakup. A nice 3rd or 4th order filter would be great when crossing that low, but the tweeter is robust enough to get by with just a 2nd order filter at 2kHz (1.5kHz or so may be even doable with a 4th order design). The clincher is that it has to be a good implementation, since the crossover happens right smack in the midrange and imperfections may be very audible because of that.
While the tweeter itself could probably work with a 1kHz shallow HPF at lower volumes, the smaller radiating area of ring radiators limit maximum usable output at very low crossover points (distortion would set in early). Doing a steeper slope in the 2kHz region would likely enhance the speaker's flexibility tremendously.
Two SLS 830668 woofers paralleled in a D'Appolito configuration (flanking the tweeter) sounds like a sweet idea, if the cabinet could be tall enough to position the tweeter correctly. The two woofers would deliver about 94dB sensitivity (half space) at 2.83V, which matches almost perfectly with the XT25BG60-04. A higher order LPF may be useful to cleanup the woofer output beyond 2kHz. There shouldn't be any need for driver attenuation, if built properly. The tweeter of course would have to be in its own chamber, to minimize interaction.
The reason I thought about this is because I prefer the sound of unattenuated drivers. Most passive attenuation circuitry on drivers tend to take away some of the openness in the sound (or color it noticeably at least), and I think it's partly because of the lack of driver damping that results from its insertion into the circuit (and/or the fact that its electrical characteristics is closer to purely resistive, thus not sympathizing with the driver's impedance curve).
Active crosses are a great alternative, but it can be just as (or even more) destructive if implemented poorly (and most people are not geared for an active layout).
This 2-way config that I mentioned earlier could yield a truly full range speaker (and a unique one, in this age of mostly smaller drivers) in the right enclosure, but the success would bank largely on how well the crossover works. Of course, a 3-way or 4-way design utilizing shallower filter slopes may indeed be easier in many cases.
