The “proper” response to a loss of data was debated a long time ago and the concensus at that time was that it was better to continue recording the last reading during what could be only a short data drop out rather than stop logging. The question is what makes the bigger mess?