In August 2015, R.S. writes:
A new report <http://arxiv.org/ftp/arxiv/papers/1508/1508.02383.pdf> from tech giant Samsung proposes that a fleet of roughly 4,600 micro-satellites orbiting Earth could solve our impending data crisis.
Perhaps they should call their idea SHRAPNEL, for Samsung High Rate Accidental Producer of Noncontrolled Encounter Losses.
The problem with LEO is that you can't see much of the ground from one satellite, so you need many satellites at high latitudes at any given time. That means the orbits are highly inclined - spending lots of useless time over empty southern oceans, and crossing the equator with high north-south velocity twice per orbit. That means the orbit of satellite X may have a descending node where satellite Y has an ascending node. If this happens simultaneously - bang! Making a whole bunch of new "satellites" in a whole bunch of new orbits -uncontrolled, and ready to party with everything else up there.
It is possible to "weave" the orbits with small eccentricities so that all descending nodes happen at a higher altitude than all ascending nodes. The problem is, the earth is oblate and lumpy, and the gravity variations modify the orbits somewhat chaotically over time. OK, fix that with thruster burns. Except that fuel is heavy, so there's a limited amount on board - and tanks leak, and thrusters fail.
So, in not too long a time, there are thousands of derelicts up there with no fuel or broken thrusters, a collision hazard for thousands of other satellites. There are already plenty of derelicts up there, and the hazard goes up as the square of the number of objects at a given altitude.
If the ground stations have small antennas, the orbiting antennas must be large, and both need to track each other, swivelling rapidly, about two degrees per second. Big antennas in space increase the collision cross section.
[ Nasty things about ignorant-optimistic popular entrepreneurs deleted ]
Server sky starts with the problems and addresses those first. Much higher non-inclined orbits, where there are only a few dozen derelicts. Ultrathin for light-sail maneuvering (the residual atmosphere in LEO creates way too much drag for this). Woven coordinated orbits as before, but closing velocities are centimeters per second, not kilometers per second. And at those high orbits, most of the gravitational anomalies are attenuated proportional to the cube of the distance.
Also, management of light pollution. Highly focused beams for better spectrum reuse. 70 GHz signalling, currently empty spectrum. Scalable to a terawatt of computation in < 100ms ping time orbits.
In the distant future, scalable to a 50 AU Dyson shell (3.8e26 watts) for computation tasks that can tolerate one day's latency. That is why I wanted those numbers; I was computing the tidal effects of the outer planets on this hypothetical shell.
Thinsat "statites" - pulled inwards by solar gravity, pushed outwards by light pressure - massing 1.5 gram/m², made of nanosystems embedded in ice, converting 0.5 W/m² meter of sunlight into computation. We may someday build a 50 AU spherical Dyson shell (1.5 tonne/km²) out of the materials in the Kuiper belt - bye bye Pluto, so long Eris, end of the argument about their status as planets.
In any case, your numbers [provided at an OMSI Science Pub a few weeks earlier] helped me guess that the tidal effects (the ratio of maximum tidal acceleration to solar gravity acceleration) were really, really small. After I got home, I plugged the numbers into a spreadsheet, and learned that the effects are 650 parts per trillion for Neptune (very briefly, as it passes underneath a portion of the shell) to 22 parts per trillion for Jupiter. A trivial contribution to positioning error. BTW, these tidal effects are proportional to the mass squared and the radius ratio (shell radius to planet orbit) to the 5th power, with a correction factor that gets large for a very close approach.
Shells like this will be imageable as highly luminous infrared disks out to 300 parsecs with the James Webb Space Telescope - obviously artificial, spectrographically very different from a cold gas cloud. The shell will be convergent technology, dependent on physics and material properties and efficient production of computation. The endpoint will be similar, regardless of what kind of intelligent life puts them there or what they use the computation for.
So, if other intelligences are out there, and we look far enough out, we will see similar infrared objects, varying in size proportional to the luminosity of the enclosed (and invisible) star, but otherwise very much alike. We can reverse engineer them from their properties, and design our own statite Dyson shell using that knowledge.
I don't know if we can learn anything about the billion-year culture inside the shell, but if it is based on some proselytizing belief system, it will probably be written in infrared albedo paint on the outside of the shell. 
Inside our shell, the real stars go away (put the telescopes outside), replaced by earth-pointed spotlights to maintain the illusion of stars (the parallax would be too high to fool the astronomers). The infrared background moves from 2.7K to 60K [ note - even this can be minimized with infrared filtering ]. Earth equilibrium temperature rises by 0.2°C . The outer "surface" of Uranus heats from 49K to 65K, though the core remains 5000K. Climate catastrophe in the outer solar system, but otherwise not much effect.
Who needs science fiction when I've got physics and a calculator?