The cyclopropylmethyl cation is one of the strange ones, comparable to the cyclopropylcarbinyl cation. Use a standard model kit to construct a model of a cyclopropane. Odd's are, you'll break your kit.
While we're accustomed to tight directional sigma-sigma overlap in unstrained molecules, cyclopropyl moieties are unable to do this. Instead there's a slight "bend" in the bonds, or better phrased, the sigma orbitals are not co-linear with the bond axis, rather they're at an angle, placing the majority of electron density off-axis.
See here:
http://isites.harvard.edu/fs/docs/icb.topic93502.files/Lectures_and_Handouts/06-Handouts/deMeijere.pdfH.A. Hart provided a qualitative explanation of cyclopropyl delocalization, J. Am. Chem. Soc. 1962, 84, 2462; J. Am. Chem. Soc. 1964, 86, 2957.
In such a case, bond rotation can place a "bent" sigma orbital co-linear with the empty p-orbital of the cation in a manner supporting hyperconjugation, shared amongst three cyclopropyl groups. At the very least, this is a somewhat rational explanation taking physical organic chemistry into account.