When the NLA uploads records they go through a match merge programme where records will match by number (local systen nos or LA control nos or LC control nos) or by title,(using the similarity algorithm. Then all records merge. If matched by number the NLA record will be the base record for the merge, but if matched by title the databse record will be the base record for the merge.
If two records are input on the same night then it depends on which file gets processed first.
All records merge. Some merge more than others. I don't like to use the word replace as I feel this gives the wrong idea
From Ian Dunn in Libraries Australia 20/8/2010