> I don't know if this happens already, but if not:
> since both scores get sorted to the same final representation, this 
> would be very possible, and often useful - just need an intermediate 
> internal merging stage to which both write, which can then be sorted 
> into time order in the usual way. The one requirement would be (e.g. in 
> the case of using csbeats) that users take care to maintain compatible 
> use of the first 5 pfields.

Yes, in general the filter command can

  1) filter only a word (macro preprocessor)

  2) interpret a score language (like csbeats)

  3) add/remove score chunks based on the input score statement,
     for example:

        if (...) add '#include "chunk1.sco"'
        else if (...) '#include "chunk2.sco"
        ...
        else if (...) add `i123 ...'
	...
        else if (...) remove/modify score statement
        ...
	etc

   4) create new chunks to be included, possibly with AI, neural network, etc

   5) send the score to csound-list and to get a correct version

   ....

And the `scsort' utility can be applied in the filter command.
I see it like a "pipe" in the unix environment (but with temp files).

tito