Where was the thread about re-using cudaHashcat.dictstat?
I've searched the archives, but couldn't find the thread about how to re-use the cudaHashcat.dictstat file made when oclHashcat has spent a lot of time in the "Generated dictionary stats for" stage.

E.g. if I have BigDictionary1.txt, BigDictionary2.txt, BigDictionary3.txt, rather than have oclHashcat re-generate the dictionary stats each time, use the saved cudaHashcat.dictstat file as an argument along with the dictionary.

(This is for -a 0 with -r rules type attacks, no Markov involved.)

Maybe this is moot, as I see the next attacks do use the info:

Cache-hit dictionary stats

Are are the dictionaries accumulated in cudaHashcat.dictstat?

I guess this is the thread, "Wordlist keyspace caching in plus?" http://hashcat.net/forum/thread-1711.html

Is there a way to tell which dictionaries are cached there?
I'm not quite sure what you're asking, because the purpose of the .dictsat file is to cache dictionary stats so that they only have to be generated once, and the file is already read on each execution. You don't need to do anything else.
Is the .dicstat file for the whole directory of dictionaries given in the command for that session?
Does it accumulate over time, or created new if a different directory of dictionaries is given?
dictstat files are created per dict.
If oclHashcat recognize any change to the wordlist it will regenerate the stats for that wordlist.