When you get this close to absolute zero active refrigeration systems are no longer effective, but the phase change of liquid helium to gas still manages to extract a little bit of energy from the system cooling it down just that much further.
But I believe that Planck was the first satellite to cool to below 1 K using only cryocoolers. Planck actually uses a dilution fridge to get to 100 mK, which is kind-of astounding.
I don't know why Hershel used liquid helium instead of a cryocooler, but my guess is that the technology for cooling with liquid helium in space is very well understood and reliable, so it's a risk thing. That's not to say that nothing can go wrong; there was a japanese telescope that lost all its helium within some very short period of time thanks to an engineering mistake.
This is a nice whitepaper about cryocooling in space: http://cmbpol.uchicago.edu/depot/pdf/white-paper_w-holmes.pd...
The cryocooler technology (say, to a few kelvins) has proved hard to get ready for space. For example, out of the 10 or so technologies that were judged most risky for JWST, the 6 K cryocooler for the MIRI instrument was the last to be judged ready for space ("at TRL 6" in the jargon) (http://www.stsci.edu/jwst/news/2007/jwst-passes-tnar).
Despite being judged ready, the JWST cryocooler has proved very challenging to build. The effort now has frequent reviews with the director of JPL (and a high-level counterpart at NGST), and tens of engineers are now working on the system.
Part of the problem, as I understand it, is that the heat has to be taken away and radiated at a site distant from the IR detector. This requires a large structure, and a deployable radiator. This large structure can't leak much heat back into the spacecraft bus or instruments, and must not be disturbed by the vibrations of launch. Additionally, vibrations of the cryocooler must not affect the telescope optics (2 micron resolution).
You can tell that these requirements are fundamentally opposed to each other ("be large, don't vibrate, don't touch anything else").
Then you either get half the mission data, or extend the mission by a factor of two, which means twice the operational labor cost, plus every other part of the craft has to be double lifetime rated, which could get expensive. Hubble used to burn thru gyros on a regular basis, so now you need them to last twice as long or launch with twice as many spares. Or maneuvering propellant if any, now you need twice as much for station keeping.
You can easily get painted into a corner where the cheapest way to run a mission twice as long is to launch two of them. At that point you're better off saying "you know the cryocool half the time and take data half the time idea? Yeah scrap that idea"
This is before we started on energy issues. A heavy helium tank doesn't use much energy. But cryocoolers on earth take quite a bit indeed, well, at least compared to a couple watt transmitter and all that. Whats heavier, a tank big enough to last mission lifetime, or a cooler and a stunning array of solar panels to run the cooler? Or since it only runs half the time as per above to prevent vibration issues, you could put a battery in which is heavy and becomes another exciting point of failure. This makes the existing power system more complicated and less reliable possibly shortening the theoretical craft lifetime to less than you'd get if you just launched a big simple tank.
Yes I know there are good reasons why they put it at a L point, but unscheduled unplanned maint is exactly what the ISS could have been perfect for. Position this dude in the same orbit but at a phase 10 miles away so its not too close, and send a live astronaut with a new battery or whatever anytime "something" fails. To say the ISS project is not managed this way would be an understatement. But someday, an unscheduled repair shop in space will exist, and probably be quite profitable.
That Planck figure is extremely impressive, mind-blowing really. That's colder than the space surrounding it.