I went ahead and did a calculation: how collisionless would "rocky" dark matter be?
rock size (m) vs. time between collisions (gigayears)
0.001 1.93013e-09
0.003 5.79039e-09
0.009 1.73712e-08
0.027 5.21135e-08
0.081 1.5634e-07
0.243 4.69021e-07
0.729 1.40706e-06
2.187 4.22119e-06
6.561 1.26636e-05
19.683 3.79907e-05
59.049 0.000113972
177.147 0.000341917
531.441 0.00102575
1594.32 0.00307725
4782.97 0.00923175
14348.9 0.0276952
43046.7 0.0830857
129140 0.249257
387420 0.747772
1.16226e+06 2.24331
3.48678e+06 6.72994
1.04604e+07 20.1898
3.13811e+07 60.5695
9.41432e+07 181.709
2.8243e+08 545.126
8.47289e+08 1635.38
That calculation is done for the Earth's "local" dark matter: isotropic 220 km/s orbits through a 0.3 GeV/cm^3 mean density. I gave it 5g/cm^3 density, somewhere between stone and iron.
Look at those numbers. If you built the Milky Way using Volkswagen-sized rocks as the dark matter, they'd last four thousand years between collisions; they'd be dust and plasma. Use 500 m asteroids, they'd last a million years before colliding and pulverizing. (Remember, these are 220 km/s collisions; they make Shoemaker-Levy look wimpy.) A 10^6 m planetoid could last for a gigayear---at least that survives a full Galactic orbit!---but at that point we're into the stuff that the EROS surveys have ruled out. Sub-meter-scale dust, of course, is not collisionless at all which is why it's never been even in the ballpark of viable dark matter candidates.
MM might object to these numbers all being so small: how has Earth survived for 4 Gy, he might ask, if Earthlike objects are supposedly so collision-prone? The answer is twofold: (a) dark matter is much DENSER than regular matter. The number density of actual rocks/planets/Earths is much, much lower than the number density I need to hypothesize to make them dark matter. (b) Dark matter is isotropic---it seems to orbit the Milky Way in a spherical halo, which implies orbits that intersect one another all over the place. The Earth (and most of the Galaxy's baryonic matter) is in the Galactic Disk, where intersections are much, much sparser than those in the halo.
Conclusion: rocks cannot be dark matter. Anything small would be highly collisional, and could not possibly remain "rocky" in a dense halo full of other rocks. Anything large would have been seen by EROS. The two ranges overlap generously.