Google commits to buying power generated by nuclear-energy startup Kairos Power

Considering that there are no commercial-scale operating MSRs, I am guessing there are some pretty significant difficulties. Like graphite pebble reactors, molten salts must be perfectly desiccated, which is impossible to guarantee under Earth operating conditions, and nobody knows what kinds of materials to use for the salt containment, or how it might be changed by a few decades of operation.

I believe it depends on the salt used and why a lot of research is looking at fluorine - floride salts don't dissolve in water and thus don't have the same risks. Another MSR design seeing a lot of attention is SSR which similarly doesn't use any chemically reactive materials. SSR is also interesting from a cost perspective as it's projected to cost half of a modern coal power station and almost a third that of a modern large scale nuclear construction putting it in line with solar plants that don't have batteries (i.e. much more cost effective).