A Williams-Kilburn tube (also called a Williams tube) is a CRT with a metal plate in front of it that relies on the secondary emission of a strong electron beam hitting the inside front of the glass envelope to store information for a reasonable period of time. I don’t know what the limits were, only that it had to be refreshed, like modern DRAM does, and that reads were destructive, like in the core memory that replaced it. Information was extracted by means of a metal plate in front of the CRT that picked up the charge differences when the electron beam hit a charge well created by a previous write (the CPU always knew where the beam was pointing).
(Incidentally, W-K tubes were the reason for the very first things that could be called computer screens: It was possible to have two tubes mirroring the same data, one tube with a metal plate in front of it to be part of the actual memory and one painted with phosphor on the inside, like a TV’s CRT, built into the front panel for the operators to look at. The Manchester Small-Scale Experimental Machine (the ‘Baby’) had this design.)
OK, it had to be refreshed but it couldn’t have required refreshing that often because the CPUs it was used with didn’t have a very high clock speed. Modern CRTs (made within the last decade or 15 years) were designed for displays only (TVs, computer monitors, oscilloscopes, etc.), meaning that persistence was very undesirable and, therefore, got designed out. It seems (I’m not sure) that a W-K tube relies on persistence for the charge wells to stick around long enough to be useful as memory.
So, if someone wanted to build a late-1940s-era computer with vacuum tubes as the logic elements and a big front panel and so on, could he use a modern CRT as the basis for a W-K tube memory?