I created this ticket as a follow-up to some observations from SERVER-113232 (see my comment). When the cache gets stuck because the page server is down and the materialization frontier isn’t advancing, WT seems to write a lot of phylog entries to the log server, even though the stable timestamp isn’t moving. I suspect it’s producing duplicate phylog entries for the same page. I’d like WT to check if these duplicates happen only when the cache is stuck, or if they can also occur when eviction is just progressing slowly. If so, should we stop producing duplicates or fail loudly by crashing Mongod?