Thanks for your help.
I found the solution in another posting. (Another case for me of working out how to describe the problem in order to find an answer!)http://pkp.sfu.ca/support/forum/viewtopic.php?f=3&t=7877&p=34327&hilit=submission+type#p34398
I'll paraphrase the story here so that it might be more findable for others with the same problem.
As soon as a TD assigns a reviewer (& before even sending the review request to them), the author can no longer see the Edit Metadata link.
The author's Edit Metadata link remains hidden until ALL reviewers have completed their side of the process one way or another. If the decision then is Revisions Required, then the author can once again see the Edit Metadata link and can therefore edit their abstract.
My 'mistake' in testing the system was in not waiting for the test reviewer to finish up before I selected the decision - and hence not Edit Metadata link for the test author.
The solution in this situation is for the TD to cancel any uncompleted reviewer assignments, then the author can once again see the Edit Metadata link (assuming the decision is Revisions Required).