Some of them can. My antenna-coupled MIM junctions worked from DC to at least 200 THz (1.5 microns in the NIR). They weren't what you'd call low-leakage devices, though. (I may get the chance to do some more this year.)
Golay cells are very slow thermomechanical devices. Bolometers are limited to the thermal time constant of the device, and it's very very hard to get that as low as 1 ns. (I have some data at that speed, using a 20-ps laser.)
Electro-optic and plasma-optic sampling has been done for at least 30 years. The very low phase noise of a good modelocked laser is a huge help, but it's pretty badly constrained by the need for the signal to be locked to the laser rep rate, rather than the other way round.
I suppose if you had a high enough sample rate, you could do what John's LeCroy does--sample like crazy and do the clock recovery in software afterwards.
Cheers
Phil Hobbs