The Lageos measurement satellites
Two passive satellites are used for laser geodesy. These are 60 centimeter, 400 kilogram solid metal spheres covered with retroreflectors, used for precisely measuring the geometry of the Earth's surface and the Earth's gravitational field.
The m288 semimajor axis (12789km) is 458km above LAGEOS 2 apogee (12330.5 km). Server sky orbits should stay above LAGEOS! Server sky orbits are not quite circular, but elliptical to compensate for light pressure drag. If we want a 2x safety margin, the minimum perigee for server sky thinsats will be 12560km, and the maximum eccentricity will be 0.018 .
These satellites are interesting because their orbits are very precisely defined, and we can use them to verify both our orbital math and learn about drag at those extreme altitudes.
\large \mu = 3.98600448e14 {m^3}/{s^2} ~ \buildrel{?}\over{=} ~ \omega^2 a^3 = 3.988154435e14 = 1.000539376 \mu
The difference is probably due to J2 perturbation dragging forwards both perigee and mean motion. However, because LAGEOS 1 is in a highly inclined and slightly retrograde orbit, the equations I have (from Piscane, page 113) suggests it should get dragged backwards. Meanwhile, those same equations suggest that M288 must be about 3.4 kilometers higher to counteract the extra acceleration due to J_2
Drag, and implications for Server Sky
LAGEOS 1 is slowed down by drag, mostly drag against the high flux of radiation particles. It is dense enough to absorb essentially all the van Allen belt particles that pass through, and change momentum as a result. The "encounter rate" is not limited by ram speed, as it is with gas molecules in the dense atmosphere, but by the speed at which the radiation particles themselves move. If \rho_p is the mass density of particles in the van Allen belt in kg/m^{3}, and v_p is their average speed in m/s, then LAGEOS sheds power into them proportional to 0.5 \rho_p v_p v^2 A , where v^2 = \mu / r and A is the effective cross section. The power reduces the orbital radius r by 0.5 M a_g dr / dt where a_g is gravity and M is the LAGEOS mass. Since dr \ dt = 1.3e8 m/s, the mass flux per unit area is:
\large Flux = \rho v_p = { M \over { A ~ r } } { { d r } \over { d t } } = 5.1E12 kg/m^{2} s
That is the radiation flux from all directions. Since LAGEOS is in a highly inclined orbit, the radiation flux changes character in a complex way as a function of magnetic latitude. At high latitudes, the flux is lower.
Thinsats spend their entire orbit near zero magnetic latitude. The flux is higher, but the thinsats present half the cross section to isotropic radiation flux  2 \pi r^2 (front and back) for a disk rather than 4 \pi r^2 for a sphere. The flux is not isotropic, as the particles are gyrating around field lines. Very complex  but as a crude approximation, assume all these factors cancel out within a factor of 2 or so.
The big question is the amount of the flux of momentum is actually absorbed. Since thinsats are so thin, much of the flux will pass right through without much momentum change. Is the fraction close to 1.0? 0.1? This needs calculation. Imagine a sandwich of thinsats thick enough to absorb all the particles. That whole sandwich will have a fraction close to 1.0, but the amount absorbed by each thinsat will be that fraction divided by the number of thinsats. So the fraction for an isolated, individual thinsat is likely to be small.
This analysis also assumes that the absorbed particles have zero average momentum. They protons actually drift towards the west, with a velocity of 700km/sec at 10MeV, 7 km/sec at 100keV, with the bulk of the particle flux at the lower energies. If most particle energies are below 10keV, there will be little effect from this.
If we assume complete absorption, a 50 micron thinsat will have a much larger area to mass ratio, and will slow down much faster than LAGEOS.
\large { { d r } \over { d t } } = { { A ~ r } \over M } Flux ~ = ~ { { 0.024 * 12789000 * 5.1E12 } \over 0.003 } m/s ~ ~ = 0.5 mm/sec = 1.6 km/year
While this analysis is quite crude, it does show that radiation flux drag will eventually bring down thinsats, far more quickly than the neutral gas density might suggest.
Orbital elements from various sources

LAGEOS 1 
LAGEOS 2 
nyo 

NORAD ID 
8820 
22195 
Int'l Code 
1976039A 
1992070B 
Perigee 
5,845.9 km 
5,622.3 km 
Apogee 
5,954.4 km 
5,959.4 km 
Inclination 
109.8° 
52.6° 
Period 
225.5 min 
222.5 min 
Semi major axis 
12,271.2 km 
12,161.9 km 
Launch date 
May 4, 1976 
October 22, 1992 
Source 
United States 
Italy 
derived 

perigee radius 
12217.0 km 
11993.4 km 
apogee radius 
12325.5 km 
12330.5 km 
JPL 

mass 
411kg 
405kg 
diameter 
0.6m 
0.6m 
eccentricity 
0.0045 

NSSDC 

periapsis 
5837.0 km 
5900.0 km 
apoapsis 
5946.0 km 
5900.0 km 
period 
225.41000366210938 min 
225.0 min 
Inclination 
109.80000305175781° 
54.0° 
eccentricity 
0.00443999981507659 
0.009999999776482582 
springerlink 

decay 
1.1mm/day 

Innovateus 

mass 
406.965 
405.38 
spin 
0.61s 
0.906s 
period( sec ) 
13246.002197265628 

\large\omega 
4.743457847e4 
Two Line Elements
LAGEOS1 TLE:
 1 08820U 76039A 11094.82079485 +.00000016 +000000 +100003 0 01141
 2 08820 109.8462 077.3556 0044210 050.5232 309.9115 06.38664784558975
LAGEOS2 TLE:
 1 22195U 92070B 11095.53472161 .00000009 000000 100003 0 3192
 2 22195 052.6462 175.4139 0138581 279.8295 078.6642 06.47294193436229
# 
chars 
description 
LAGEOS 1 
LAGEOS 2 
1 
0101 
Line number 
1 
1 
2 
0307 
Satellite number 
08820 
22195 
3 
0808 
Classification (U=Unclassified) 
U 
U 
4 
1011 
International Designator (Last two digits of launch year) 
76 
92 
5 
1214 
International Designator (Launch number of the year) 
039 
070 
6 
1517 
International Designator (Piece of the launch) 
A 
B 
7 
1920 
Epoch Year (Last two digits of year) 
11 
11 
8 
2132 
Epoch (Day of the year and fractional portion of the day) 
094.82079485 
095.53472161 
9 
3443 
First Time Derivative of the Mean Motion divided by two 
+.00000016 
.00000009 
10 
4552 
Second Time Derivative of Mean Motion 
+000000 
000000 
11 
5461 
BSTAR drag term (decimal point assumed) 
+100003 
100003 
12 
6363 
The number 0 (formerly "Ephemeris type") 
0 
0 
13 
6568 
Element number 
0114 
319 
14 
6969 
Checksum (Modulo 10) 
1 
2 


1 
0101 
Line number 
2 
2 
2 
0307 
Satellite number 
08820 
22195 
3 
0916 
Inclination [Degrees] 
109.8462 
052.6462 
4 
1825 
Right Ascension of the Ascending Node [Degrees] 
077.3556 
175.4139 
5 
2733 
Eccentricity (decimal point assumed) 
0044210 
0138581 
6 
3542 
Argument of Perigee [Degrees] 
050.5232 
279.8295 
7 
4451 
Mean Anomaly [Degrees] 
309.9115 
078.6642 
8 
5363 
Mean Motion [Revs per day] 
06.386647845 
06.472941934 
9 
6468 
Revolution number at epoch [Revs] 
58975 
36229 
10 
6969 
Checksum (Modulo 10) 
5 
9 
. . . 11094 = April 4, 2011 . . . 11095 = April 5, 2011
References: