One of the problems with refractory materials is that they tend to be brittle at room temperature. The ductile-brittle transition of a material, if it exists, tends to be not too far from its melting point, so materials that are ductile at room temperature tend to not survive 1000° or 1500°, with a few exceptions.
As a result, when they heat up and cool down, they tend to crack.
The Neolithic solution to this was to pit-fire pottery: you bury it in a pit and throw burning coals on it, then, perhaps, cover up the pit, partly or fully, with dirt. It’s fine if the dirt is brittle and cracks, because it’s already powder; you aren’t demanding any strength of it. If it crumbles, it crumbles onto the other dirt and pot underneath it.
A modern equivalent of this is the salt bath or sand bath used by opticians and labs to heat up materials to a desired temperature, either to partly melt or soften them, or to provoke some reaction.
It occurred to me that you can do this with heating elements in the salt or sand, thus achieving a high-temperature capability without constructing a castable refractory with any kind of strength. Quartz sand is cheap, and olivine sand isn’t that expensive. You can jam temperature sensors into the sand too, and they might be able to be at a significant distance from the thing that’s heating up, so they can be at a lower temperature.
Vermiculite, charcoal, ash, cat litter, and plaster of paris may also be useful; they are somewhat refractory substances that thermally insulate better than sand. Even perlite should be useful up to a point, and that point is about 900°.
Higher temperatures may require the use of arc heating rather than solid heating elements, although carborundum heating elements can extend this considerably.
Carborundum itself was discovered in just such a way and is still produced by this process: a sort of arc furnace is set up under a layer of silica sand with carbon electrodes, originally in an iron crucible, but nowadays at a much larger scale.
To some extent you should be able to measure the temperature distribution within the pile of fluff with temperature sensors that are not exposed to its hottest part. You can estimate the conductivity, thermal mass, thermal resistance to ambient, and heat input using a few sensors at known or estimated locations, a continuous measurement of the thermal input, and some PDE solvers. This could potentially permit very precise control of the temperature distribution within the pile.