I think this is as a result of drift of the kernel drivers (which are actively maintained) with respect to the userland raudutils (which haven't been touched since 2006), but, in any case, the "raideng" component of the raidutils is silently segfaulting when starting up. I traced the issue to the BuildNodeNameList() function, which is responsible for identifying the driver's device names but, before spending more time on it, I identified a Debian patch that addressed multiple issues, including this. The new Debian-derived patch replaces the raidutils-0.0.6-misc-fixes.patch, which seems to be an earlier version of the Debian patch.
Created attachment 202055 [details, diff] Patch to fix reported issues.
Created attachment 202057 [details] New ebuild applied the above patch.
Created attachment 202063 [details] My emerge --info.
Created attachment 230147 [details, diff] Patch to current raidutils--0.0.6-r1.ebuild to correct issue. Per a recommendation from Polynomial-C, I should be providing patches to ebuilds as opposed to new ebuilds to make it easier for the developers review the changes. Also, since the bug was filed, some cleanup has been done by Ulrich Mueller to the in-tree ebuild. The patch takes these changes into account as well.
This looks to be fixed by 0.0.6-r2 (it has latest Fedora changes).
This appears to work on the hardware. My RAID management scripts will be exercising it more or less continuously. Just out of curiosity, why was this pushed to stable? Why isn't it in testing? - John
Reopening for comment on skipped testing phase. - John
Well, maybe nobody else uses this but me. It's been stable on my system for a looong time now. Closing as FIXED.