Currently, if we get a group-refresh for data not in our PendingWrites we bail out. We assume that there should not be a situation that we don't have it in PendigWrites, nor in chunk store, nor in ongoing gets/data-holders. Investigate if this can be and see what the solution is. With MData even if one gets a refresh for an entry which is not in any of these places, we cannot simply ask for that entry (that would work, however) as there might be other entries that are out of date if such a situation was hit and there is no way to know without asking for complete MData from every one in each such case (which ofc generates a lot of traffic).
Currently, if we get a group-refresh for data not in our PendingWrites we bail out. We assume that there should not be a situation that we don't have it in PendigWrites, nor in chunk store, nor in ongoing gets/data-holders. Investigate if this can be and see what the solution is. With MData even if one gets a refresh for an entry which is not in any of these places, we cannot simply ask for that entry (that would work, however) as there might be other entries that are out of date if such a situation was hit and there is no way to know without asking for complete MData from every one in each such case (which ofc generates a lot of traffic).