I don't think TRPoW is a recipe for disaster. Far from it. Keep in mind that Bitcoin worked with small group of miners for many years.

As I wrote to you previously. We can find an optimal value of N2 that does not undermine the security of Bitcoin with TRPoW. Intuitively an upper bound of such value is O(SquareRoot N1). Assuming that the winner of Second Round is among N2 List of N2 miners instead of only N2 first miners. I think an optimal value of N2 could be in O(Log(N1)) based on a similar calculation done by EPFL Distributed Algo team: https://arxiv.org/abs/1908.01738

You can also see Bitcoin with PoW as Bitcoin with TRPoW where N2 = N1.

If we assume for the sake of argument that N1 = 1000000 and N2 = Square Root of N1 = 1000.
The reduction of energy can go drastically because we will have 1000 000  - 1000 = 999 000 miners (99.9% of the network) not doing the Second Round. We will save huge amount of electricity without undermining the security of the blockchain.