64 bit Truncation

Yes, going from 64- to 24-bits is wordlength reduction. Dither basically takes care of rounding errors when you do that. It has nothing to do with SRC, which you don't need to perform anyway. 
 
When you export from SONAR, select 24 bits and Pow-R dithering and go. Don't worry about the 40 bits you're discarding.

Yes, conventional wisdom is to only dither once.
 
Your 64 bit file, to my limited knowledge, cannot truly exist as an audio file (I don't know of any interface that would play it). As you say, the internal math computations are happening at 64 bit but the resulting audio file should be 24 bit.

Are we sure we are talking word length here and not the processing 64 bit? 
Sonars default for processing is 32 bit and the option of upping that to 64 bit. 
But recording bit depth for every interface I read about is either 16 or 24 bit. 
What interface do you own that has a 64 bit recording word length? And if so what is gained buy such a crazy setting?? 0.00009%?? 
 

That the interface can't record or playback 64 bit has nothing to do with it. You could save an audio file to 128 bit or 1024 bit for the purpose of editing, this will (at least in theory) prevent rounding errors when you use fx or bounce the file or change gain or any such thing. Upon playback of course the interface will play it at 24 bit but that doesn't mean the audio file "can't exist" at 64 bit. In fact the standard bit depth is 32 in Sonar I believe.

Dithering to 24 bit really doesn't matter:
 
1. First of all, stuff at the level of 24 bit dither isn't remotely close to being audible under any real world conditions whatsoever.
 
2. Even if it was, the 24 bit dither would be buried under other noise - the best 24 bit converters can do is equivalent to 20 bits or so, so the dither is always buried under the noise from the converters in addition to other noise and it's all below what your ears can hear anyway.
 
3. Your audio almost certainly already contains nothing but noise at the same level as the 24 bit dither, which means you have "self dither" and dithering thus provides no benefit because the noise in your signal is already doing the dither's job.
 
4. You don't need to worry about repeatedly dithering causing noise buildup, because of #1 and #2 and the fact that you'd have to apply 24 bit dither literally dozens of times before it could possibly matter.
 
 
IOW 24 bit dither really just doesn't matter at all. You can dither or not dither or dither repeatedly at that 24 bit level and it will make no difference whatsoever.
 
You can think of it as changing the font on a street sign 2 miles away to make it "more legible" - it might make a difference from a distance that's just at the limits of your vision (like dither with 16 bit audio), but from miles away it's beyond irrelevant.