Thanks for that, Mach.
Did find those papers, and they do mention the advantages of challenge-and-response, i.e. redundant checking and thus having 2 chances to spot an error.
FSF paper on error trapping was good as well, but did not go into benefits or draw-backs of read-and-do, although it made some good points on flow execution.
My main problem with read-and-do for normal checklists is when task gets crossed with reading, i.e. when the task is in the area of responsibility of the pilot reading the checklist. Chances of an actual cross-check being performed correctly are low, I feel - just need the data to back it up...