Is it possible to use thermal byproduct of daylight operation to power night time operation of volatile computer storage?
In my setting, a moon-like moon with almost no atmosphere has one of its sides (the far side) populated with nodes powered with sunlight (solar flux of said planet is around $1247 W/m^2$), and day-night cycle is 27 days long. The computer storage on this node is volatile, meaning it needs continuous power supply to keep it from losing its memory. The node's storage requires constant power at $4\times10^6 W$.
Each node operates at $2\times10^7 W$ of power, easily powered by sunlight during 13.5 days long daytime. About $4\times10^6 W$ of it is deposited in flywheel energy storage during daylight, that is 13.5 days. Naturally, each node radiates $2\times10^7 W$ as thermal energy, so I equip them with radiators that operate at 600K temperature, with coolants exiting at 400K, radiating at a rate of $7656 W/m^2$ according to equation on this answer.
As mentioned above, at daylight the radiator outputs coolant at 400K temperature, and this heat is stored on water reservoirs, for the sake of this analysis assume that the reservoir is near-perfectly insulated. This reservoir stores $3.52\times10^8$kg of water, that stores about $1.87\times10^14J$ in the form of heat.
Now, at night the node is allowed to operate at lower temperature, let's say it could be as low as 273K, and with hot reservoir at 400K, this is 127K of temperature difference. Again as in linked answer, the temperatures of the hot and cold reservoirs of your power generation system define Carnot efficiency, and in this case it is ideally at 37.5% efficient. But let our heat engine works at 35%, this means from the stored energy only ~$5.6\times10^6W$ of usable work could be extracted for the entire night time (13.5 days).
Flywheel energy storage charged from 13 days of continuous operation at $4\times10^6W$ during daylight is now extracted, and due to limited conversion efficiency at around 80%, only $3.2\times10^6W$ of power could be extracted.
In total, energy from heat engines and flywheel energy storage provide us with $8.8\times10^6W$ of usable energy during night time, enough to power our computer storage, and some energy is left for node maintenance.
Now, this system also radiates energy at around ~$2\times10^7W$ (heat from reservoirs and from energy generated by extracting the flywheels). The radiator at night time took 400K coolants and outputs it at temperature of 250K, according to linked question the radiator would radiate heat at a rate of ~$1400W/m^2$.
My concern is that although the configuration appears realistic enough to be workable for me, this results in total of constantly radiated heat of $2\times10^7W$ during day and night, despite the fact that each node is powered by $2\times10^7W$ of sunlight only in daylight. At first I thought it is possible because of different working temperature, 400K during daytime and down to 273K during night time, but I can't be sure.
Therefore, I came up with this question: Is this configuration of utilising thermal byproduct to power night time operation possible? If it is not, then where is my error?
[This question was on sandbox]