I am using Mendeley for managing my papers. I like it because of the reference manager and the multi-platform/OS. I have approximately 400 papers that weigh almost 380 MB, all of them stored in a Dropbox "Papers" folder. Inside that folder I have subfolders like "Theme A", "Theme B", "Theme C" in order to get quicker access when I manually search for something. I chose this setting because I can upload to Dropbox from any machine and I thought it could be a better idea than space provided by Mendeley (I still don't have a sound reason for this choice). So far, so good. I have configured Mendeley to use the "Papers" folder as a watchfolder so every file I put in there gets synced to Mendeley. Every time I want to download a paper I just throw it into one of the subfolders and voilá (I like this because papers often come with counter-productive titles, often with authoryear or author_year or whatever).
I have noticed that every single time I log in to Mendeley Desktop (from my laptop or any of the Desktop computers I use (work + home)) duplicate files are created. Dropbox files are not getting duplicated, however, Mendeley files keep duplicating forever. I want to avoid searching for duplicates and merging or deleting files.
I have asked other people that use Mendeley but they do not sync from dropbox as I do so they do not face this problem.
I noticed that most of duplicates have to do with different information (e.g., doi, PMID, year, abstract ,etc...) that Mendeley is not getting right the first time it loads the article and I have to manually ask for it to search (which it does marvellously).
I am about to migrate my data (new computer/job/etcetera) and seems to be the right time to learn to do the setup correctly or change to Readcube or another bibliography manager.
UPDATE
I have migrated my data. I currently have only one computer and switched strategies. I read outside of Mendeley because pdf readers keep comments with better compatibility that Mendeley (also not everybody has Mendeley nor reads on it). As November 2016, I have a 'papers to read' folder in which I download papers and a 'stored papers' folder that I will never ever touch again (read modified papers go here and here's from where Mendeley gets the references). This approach is still not perfect (i. e, Remember to move files after reading, When you want to cite something that is currently in the 'papers to read' you don't have it, Mendeley/Pubchase recommends you papers you already have in the 'papers to read'). I am not ready to go for a desktop computer and I might not want to have Mendeley installed there if I buy such computer.
This must be intrinsic of Mendeley's functioning, meaning hard coded on purpose. We must remember that Elsevier's business is now to sell you cloud space for storing your papers. Hence, they are not happy with you paying Dropbox or anybody for a much more reasonable price per Gb. As always, their profit margins are all that matters.