I did some quick searching and a tiny bit of experimentation with the tech designer in-game, and it appears there are no longer any efficiencies of scale that come from larger reactors; the only one that used to exist, internal armoring, no longer does. For example, a Size 1 Stellerator Fusion Reactor with no power boost results in 2 crew requirement for a cost of 18 Boronide and internal Hits To Kill (HTK) of 1. A Size 2 Stella-blah reactor requires 4 crew and costs 36 Boronide, but also has an HTK of 1 - twice the size, twice the cost, twice the fuel, twice the power output, but still just as durable. A size 10 takes 20 crew, costs 180 boronide, and has an HTK of 5. On the other hand, anything smaller than a Size 1 reactor (0.1-0.9 HS) will see everything else scale linearly, but the HTK drops to 0: if hit, the component is destroyed, but the shot keeps on going to hit something else as well.
In other words, if I'm parsing this right, your ships actually become *easier* to kill if you use any power reactor size other than 1: Hits To Kill scales by half the reactor size in HS, rounded...kinda weirdly, actually (Size 3 has an HTK of 1; Size 4 an HTK of 2, but it's not a simple halve-and-truncate because Size 1 has an HTK of 1 rather than 0). Smaller reactors are useful in order to provide granularity without devoting unnecessary space to power and larger reactors will keep your ship fully powered for longer if they are damaged but not destroyed (reactors continue to provide full power until destroyed), but otherwise, you're better off building an array of small reactors than a single massive reactor. Reactor explosions for smaller reactors are also smaller as well, as the explosion size scales proportionately to power, but since explosion chance should be calculated on damage rather than destruction, there's no difference between having a large reactor damaged multiple times or several small reactors destroyed in succession save for the fact that the single large reactor won't absorb as much damage as the several smaller ones and thus won't roll the odds as many times. Arrays of small reactors will also save on research time, since the research cost of larger reactors also scales linearly.