Regardless of how the recreate got started (it really shouldn’t be doing it automatically) recreate performance is s known issue.
Much of it seems to be related to inefficient SQL checking the database for already existing records. While I found some time to build some metrics for my specific test case, I haven’t gotten around to actually testing any fixes. 
What generally seems to happen is the progress bar moves along until about 75% then seems to stop. Duplicati of actually still working, but it takes progressively longer to finish a percent of work the further along the process is.
In regards to recreate vs. start over it MIGHT be faster to start over, but you’llv likely use upload bandwidth re-pushing the files. You’ll also have a break with the history of previous backups.
One thing I’ve recommended to people is if they want to start over point to a new destination folder and leave the “broken” one in place. If needed, a direct restore from the “broken” destination can be done.
That way you’re not left with NO backup while the new one is “filling”. And of course you can keep the “broken” destination around for as long as you care about the history it might contain.