I've seen a couple of posts on this, and it seems pretty much to be a non-issue. There's only one way I've come to "fix" this, and you may try it yourselves. Rebooting doesn't seem to fix it, and honestly that would be a lazy way to fix it to begin with, IMHO. Sometimes they can recover, not often. I see it mostly when one of my Frontier Editions start a task, and when another picks up on the task, either reboot, restarting the application, or perhaps when the task restarts for whatever reason. This works about 50% of the time, and when it doesn't, I typically just abort the task. It usually happens when the task is within the first 5% of completion, so you don't typically lose much of the crunching. For my setup, I have soft links to both the BOINC directory and the setiathome.berkeley.edu project directories in my home directory, which is where I use this "fix" or workaround. I suspend the task, and get the properties from it. Two pieces I need from the properties page are the name of the task, and the slot directory.

bash-3.2$ ls setiathome.berkeley.edu/blc56_2bit_guppi_58693_09214_HIP98819_0144.2159.0.22.45.255.vlar*
setiathome.berkeley.edu/blc56_2bit_guppi_58693_09214_HIP98819_0144.2159.0.22.45.255.vlar
setiathome.berkeley.edu/blc56_2bit_guppi_58693_09214_HIP98819_0144.2159.0.22.45.255.vlar_0_r940687828_0
bash-3.2$ rm setiathome.berkeley.edu/blc56_2bit_guppi_58693_09214_HIP98819_0144.2159.0.22.45.255.vlar_0_r940687828_0
bash-3.2$ rm BOINC\ Data/slots/23/*

After these files are deleted, I resume the task, and let it run.

EDIT: I noticed this a lot more when I was running an RX 580 along with a Vega 56, so it appears that dissimilar, though comparable, cards may have some bearing on why this may be happening, but that is just conjecture on my part. Just some ideas for those of you experiencing this issue. I've also noticed a higher correlation with higher SBS values. If you're getting this a lot, decrease the values and see if that doesn't decrease the occurrences.

---