No, it makes sense and can remain a hex. r = sqrt(A/pi). Then round up, that's good enough. Might need to tweak sensor upgrade scales because it is appropriate gameplay-wise to have a large sensor radius in the mid-late game.
Well, for starters, your formula doesn't work on the hex grid, at least not under the rules we have. The area revealed by a sensor of range R is A = 3R^2 + 3R (+1 if you include the tile upon which the sensor sits), and so (ignoring the tile in which the sensor sits) you really ought to have R = -0.5 + sqrt(9 + 12A)/6 for the sensor range.
Beyond that, without breaking out a calculator, tell me how much range you'll get with a sensor area of 10, 100, 500, 1000, 2000, or any other arbitrary number. You don't know, do you? That is a major reason why you should not provide sensor effectiveness in terms of area revealed. Especially since sensors have two primary roles - early warning and surveillance - and only one of those cares at all about area revealed. The amount of warning time you get is a bit less than linear with sensor range (since we're dealing with discrete time rather than continuous time, there's a floor(X) function involved; any amount of warning less than a full turn, or which doesn't bring the warning time to the next higher integer number of turns, doesn't really do anything for you), and surveillance targets are at best roughly linear with area. Exploration rate, meanwhile, is linear with sensor range under the assumption of constant speed (and can in fact be reduced by increasing sensor range), except potentially on the first turn you rushed a sensor ship out (and therefore probably had a lot of unexplored tiles close enough to the point where the sensor ship emerged to suddenly gain a revealed-area-dependent amount of information). Why exactly is area a good way of providing the player with information on the quality of their sensors?
The good news is that they don't have to, they just need to understand "more sensors => further sight".
Problem is, if sensor components reveal a set amount of area per sensor, more sensors does not necessarily imply greater sensor range. There is no constant area such that adding another sensor component guarantees that the sensor range will actually increase since we're dealing with discrete ranges. There is, in fact, a great deal of potential dead space, where adding several sensor components will in fact not accomplish anything at all.
Moreover, tiles revealed is a very poor metric of the quality of the sensors. What do you care most about with sensors? How much time you have to respond to an oncoming hostile target? That amount of time is dependent upon sensor range (specifically, advanced warning time is greater than or equal to max(0, floor((R - M + 1)/M), where R is the effective sensor range at a specific point and M is the number of moves per turn the hostile target can make). Where does area come into this? Nowhere directly. Exploration? Well, the turn your ship is created you could perhaps reveal up to 3R^2 + 3R + (2R + 1)*M tiles, but after that it's no more than (2R + 1)*M tiles per turn, where M is the number of moves per turn of the exploration vessel. Where does the area revealed by the sensors come into play here? Once again, nowhere directly, except possibly on the turn your ship is created. Surveillance? Now here's something that is, to a degree, dependent upon area; the maximum number of targets which can be observed should be roughly linear with area on a map with a more or less uniform distribution of surveillance targets. On the other hand, the safety of the observer is dependent upon its ability to observe from further away than the observed can strike, and its secrecy is dependent upon the ability of the observer to watch from beyond where the observed can see. Beyond that, just how much is area actually worth for surveillance? Of those several hundred to several thousand tiles my ship can observe, how many do I actually care about?
So, of early warning, exploration, and surveillance, the three primary functions of a sensor ship, we have two which predominantly range-dependent and one which is somewhere between range-dependent and area-dependent. Why, exactly, is area a good metric for indicating how much I'll get out of a sensor component?
Just use the sensors as they were and add a small negative percentage to each one and be done with it. Now you can't stack them up to infinity and still get a decent amount of range without being OP in the process.
I regard your proposed solution as being worse than the problem which it is supposed to remedy. -N*X% sensor range with N sensors, without the penalty even ignoring the first component? Reduced mass efficiency, maybe, but not reduced sensor range per sensor. Yes, for certain types of detectors, you can make a realism-based argument for it; for others you cannot, however, and it's not like Galactic Civilizations sensors behave like anything currently in existence anyways.
Besides which, you already cannot stack sensors up to infinity. You're being excessively hyperbolic, don't you think?