Vultr finally resolved the problem this morning, 49 hours after I initially reported it.

"We found and issue on one of the upstream switches that had a failing process. We have corrected the issue and now I see no more errors accumulating."