I see there’s a new 2016 study of the DOAJ in New Library World (Vol. 117, 11/12, pages 746-755). The researchers found that in the DOAJ…
“roughly 20-25% of the [journal homepage] URLs redirected to another URL” but that “only 2.11% of 9,073 journals [proved] to be inaccessible”
… once the redirects were followed.
Two automated tests were done (using home-brewed Excel wizardry, rather than dedicated linkbot software) of all 9,073 titles, one month apart, pinging each journal’s homepage. They followed this up with a manual check on all the URLs of the still-inaccessible journals.
The research seems to have been quite thorough, although I’d observe that a homepage URL is far less likely to be broken than the deeper direct article URLs on the DOAJ’s table-of-content pages. Article page / PDF URLs can be easily broken, for instance by the journal moving from WordPress to OJS or visa versa. A similar test might usefully be run on a sample of DOAJ article URLs, although I must say that I haven’t noticed any problem on the DOAJ in that respect.
I see that Bentham Open (aka Bentham Science Publishers, not directly indexed in JURN) provided 67 of the inaccessible titles. For some reason they are still in the DOAJ after the recent purge, but my quick tests on the DOAJ’s Bentham URLs found all those tested to be unresponsive. That was last night, and they were tested again today and again found to be unresponsive. So I’m not too worried about their popping up in JURN results (via the DOAJ indexing) and I presume that the DOAJ will have them out fairly soon for 404-ing.