As has been said above, the main problem is the filtering of slow objects from display, but regarding radar reflectors, this document gives a good background on retroreflectors:
http://www.dtic.mil/get-tr-doc/pdf?AD=ADA456839
It might be possible to make a very light and cheap trihedral (box corner) reflector out of, say, carbon rods and aluminium film, but it'd be very bulky. This has the advantage of being simple and working on any radar frequency.
But, given that civilian primary radar operates on the manageably narrow band of 2.7-2.9 GHz, where compact high performance planar antennas can easily be built, it would be possible to build small Van Atta reflectors, that would basically look like PCBs. In the above linked report, a Van Atta reflector for 10 GHz radar, approx. 30 cm square and 5 cm thick, had a radar cross section of 100 m^2. For comparison, this is about the same as a 747, and 100x more than a light GA aircraft.
At 2.7-2.9 GHz, and with two reflectors the size of, say, iPads, mounted back to back to cover most incident angles, should give at the very least as much RCS as, say, a microlight. Such reflectors could in principle be made as regular PCBs cheaply (say, <£100 for the whole setup), or could be made smaller by incorporating active amplification electronics (fairly simple and cheap at these frequencies). But, it would be difficult to mount to a paramotor in the correct orientation to give useful results, and would require highly specialised development, expensive testing, and the cooperation of NATS and/or the CAA, and Ofcom if it incorporated electronic amplification, thus making it end up very expensive.
Transponders are the more realistic solution, just as soon as someone makes a handheld battery powered one (like a handheld transceiver).
