Actually not necessarily. Many light, cheap and small antennas can also work.

See p.8 and 10 for those used in the current state of the art telescope at 160MHz https://www.aanda.org/articles/aa/pdf/2013/08/aa20873-12.pdf And https://cdr.skatelescope.org/#photos?lfaa for those used in the SKA, one of the largest future radio telescopes that is still under construction.

Getting the distances between the telescopes is actually easier in space. Both distance measuring and datatransmission could be done with lasers when there is a clear line of sight. Moreover, once an orbit is established, the laws of Kepler are 'followed' and predicting their mutual distances is something we can do extremely well. On Earth with very long baselines it is much trickier and things that need to be taken into consideration are cablelength differences due to temperature changes, tides and continental drift. (Continental drift is actually measured with radio telescopes: in the reverse problem when the location of a set of sources on the sky is known to high precision one can establish at what speed the distances between the telescopes is changing.)

You might want to look into what the NRO is already putting in space (supposedly) [1]. If the reports are to be believed, 100m dishes are already there, just pointing at the earth instead of into deep space. To your second point, it seems that relative positioning is mostly a solved problem at least for satellites in GSO (that's a core part of how GPS works). My personal opinion is that space-based radio astronomy is mostly a problem of cost, not of available tech.

[1] https://en.wikipedia.org/wiki/Orion_(satellite)

Not a solved problem, GPS is too inaccurate for interferometry. The problem is you want to do correlations between radio signals; consider a 1 GHz signal, which has a 30 centimeter wavelength. Error compounds as you integrate longer and longer signals. To do anything decent, you need to know absolute position within millimeters, and it can't be drifting too much.

Also, the positioning error isn't the whole story; there's also timing drift. Rubidium GPS-disciplined oscillators actually drift too much instantaneously to be useful...

Laser interferometry will pinpoint the distance to far less than a millimeter.

Military spacecraft are almost always ahead of scientific ones, and a 100m foldable dish might be an engineering marvel, but it's a bit different from a 100m telescope. An actual telescope needs to guarantee that your data are correct within certain tolerance, and not just good enough to get a good SNR from another spacecraft.

Besides, as another poster noted, VLBAs need extremely precise positioning which is just another rabbit hole. Spektr-R orbit determination was really tricky for its baseline of 300000+ km.