So, in our world of programmable analog circuit arrays and perfect rail-to-rail opamps, when would you *need* to use a tube? When does using silicon not make sense?
I've been fortunate to have worked in a wide variety of fields, including radiation monitoring. Some of the sensors provide signals in the microvolt/nanoamp range, and local gain is needed before forwarding the signal elsewhere. But when the sensor is near the center of an operating nuclear reactor, gain can be tough to find. How to get local gain?
Specialty tubes to the rescue! A vacuum is naturally radiation-resistant. But careful selection of materials is needed to make a tube that will have a useful lifetime and still work at high temperatures in the presence of massive levels of ionizing and neutron radiation. It turned out that minor process tweaks to existing tubes did the trick, and a dual-stage high-gain amp provided all the gain needed. (Our work was also used on satellites operating in the van Allen belts.)
What finally killed tubes in this application (decades after tubes disappeared everywhere else) was their accelerated aging. Inside a nuclear reactor, a lifespan of a 7 years often isn't good enough.
So we had to find ways to get the raw sensor signal away from the reactor to external amplifiers. The cost was moved from the tubes to ultra-low-loss cabling, and the signal was fed to what was basically a logarithmic nano-ammeter.
Fun all the way, no matter how you slice it. The conversion to a nearly all-digital system (reducing the circuitry between the ADC and the sensor) was boring in comparison.