It's related to the MPPT trackers. Basically what is happening during these early morning scenarios is open circuit voltage will initially measure beyond 140V, which turns on the system. However, right when the panels experience load with very little sunlight available, there will be a voltage drop, which drops string voltage below 140V, cutting it off. Cycle/repeat, until the panels can produce enough amperage to keep the MPPT trackers happy, which in my case would be somewhere above 300V. This phenomenon is very similar to attempting to start a car with a degraded battery - it may measure 12V at first, but the moment the starter relay is engaged, will drop voltage significantly while simultaneously not providing enough amperage to start the car. A "good" battery will still experience some voltage drop, but will also provide enough amperage to power the starter motor as needed.
Regarding lost production between 140V and 300V - in theory, yes. In practice, no... or very little. Inverters are not 100% efficient, and there is a baseline amount of energy required to keep the inverter circuitry itself powered on. Any plain old inverter (think 12 => 120V) will have some level of parasitic drain of 10's of watts just to remain 'on', even with no load attached. This number is typically higher on larger inverters... especially when approaching the scale of 10kW. If I had to choose between drawing 100W from the grid or battery, to capture 10-20W from the panels, I'd rather cut the panels and wait until they are ready for more production.
As for floating between 300V - the voltage output is not linear, and will very quickly reach peak, even with relatively little sunlight. Technically, this should be close to the Vmpp value provided by the solar panel datasheet. It is a balance between Voc (which is what the inverter first sees), Vmpp (voltage the panels are rated to be most effective, see datasheet), and the MPPT trackers, which constantly adjust load in an attempt to hunt for maximum power generation, for any given moment in the day.
Even with low current, the voltage is still high, and there will be an arc for every engagement/disengagement cycle. This will wear down any mechanical relay over time, so it's better to not do this excessively. Totally my fault for not setting this sooner or monitoring the behavior closely, but I can see how this will absolutely cause certain types of premature failure on any inverter.