Check out org-transclusion, it covers some of the use cases you describe or could be used to implement them. The issue with giant results, especially those that have very long lines, is more challenging, but being able to specify that different output streams should go to files instead of results might be a start, or having them go to a file and having process do the equivalent of tail on the file to continually update the results stream of the buffer would be really useful (if it could be done without disrupting other editing flows. The shortest path to splitting file vs buffer would probably be to decorate org-bable-insert-result (in ob-core.el) to detect the size of the result and write the full result to a separate buffer if it is beyond the limit you set. Best, Tom
https://github.com/nobiot/org-transclusion