Copying them into the database is less secure because they’re stored in just one file. E.g. if there’s a power outage or OS X freezes/panics and you have to reset your computer, the filesystem might be corrupted afterwards. And if the file containing the PDF documents is damaged, all your documents might be lost.
But if the documents are stored as separate files, it’s not very probable that this will happen - those PDF documents are never modified and there are multiple files.
However, this doesn’t matter if you have a good backup strategy (e.g. burn your data regularly on a CD/DVD). And a backup strategy is highly recommended to everyone: Computers, harddiscs, filesystems, processors, caches, memory, software, the operating system - everything might fail. And this does actually happen - we’ve experienced this multiple times during the last years.
I’ve got thousands of PDF files covering environmental science and engineering as well as a variety of other topics. Many of them run hundreds of pages in length.
My own preference is not to load the PDF files themselves into my DEVONthink database, but to import them as text or RTF files so that they can be searched by DEVONthink. The new Index import mode will be my choice from now on, for most large PDF files. (Exception: I’ve been working with a batch of encrypted PDF files recently. Index import doesn’t work with these, but I’ve found that TextLightning 3 does a nice job of capturing text from these files.)
Like many others, I prefer reading PDFs in Preview (and prefer Preview to Acrobat).
As Christian says, database maintenance and backups are good!