looking into the log file, the nearest i can trace down the "upkeep loop" breaking down is this:
NullReferenceException: Object reference not set to an instance of an object
at DungeonWorldEventManager.TriggerEvent_OnPregnancyEnd () [0x000f8] in <f2417dd6da484ffa8597b424d99bbdbc>:0
at DungeonWorldEventManager.TriggerEventSet_OnNewTurn () [0x00000] in <f2417dd6da484ffa8597b424d99bbdbc>:0
at DungeonManager.NextTurn () [0x001e1] in <f2417dd6da484ffa8597b424d99bbdbc>:0
at DungeonManager.Update () [0x00046] in <f2417dd6da484ffa8597b424d99bbdbc>:0
in this particular instance, the upkeep loop breaks during a birth. the birth may or may not be blocking the return message of servants i have sent out of the dungeon. of note, as i've been reloading a save constantly to narrow down the event breaking the loop, an invasion once blocked the problematic birth, but upkeep was broken all the same. this is why i suspect there is a competing "return from outside mission" event being blocked by the birth event. and to be specific, the servant returning is not the same servant who is giving birth.
the log is also full of an "amical behavior" fail, which may be related to the servant being born and stuck at zero mind because upkeep breaks before the child is fully generated and placed in the game world.
there's a lot of "failed to play audio source" errors, which i suspect might be the zero-mind newborn's sobs.
there's a lot of geometry errors, particle effect errors, and navmesh errors seemingly tied to weapons.
finally, in the end-of-file memory scan, there's a remarkable number of logged errors in what i assume to be multithreading reports.