Thank you. The links were the same (present and pointing to the same location) on all
three nodes (I had already looked through the possibility of a split brain following some
other suggestions in older emails on this list).
Sometime late Friday I found a message that suggested I should place the cluster into
global maintenance mode and run `hosted-engine --reinitialize-lockspace`. After doing that
it seems to have recovered (I presume that command removed/regenerated/fixed the link).