Such a "super sensitive device" might be the bog standard 1210 MLCC. The manufacturer may explicitly forbid any hand soldering activity completely, and gives very strict guidelines not only about board flex, but also about the footprint, max number of reflow cycles (such as only one), and the reflow profile.
I have had a Murata 1210 MLCC die after a bit careless hand-soldering operation. It died after running for 1000 hours no problem, then sitting unbiased in humid summer air, during the next power up. It has became a low-value resistor (high enough not to blow properly sized fuse, and not to vaporize instantly) - glowing bright yellow, nearly starting a fire. After looking at the matter, indeed, the manufacturer forbid the hand soldering completely. (The actual mistake I made was to solder multiple capacitors too closely together, in a shared copper fill with solder mask openings, without thermals. But, there was no mechanical stress involved, which is the thing most people seem to think is the only issue with MLCC's).
I know everybody here says "nah, haven't happened to me, it's not a problem". But it may be a problem. Follow the standard advice for MLCC, especially with 1210 and up.
Yes, preheating MLCCs to reduce thermal shock might indeed make some sense, and may not be overkill. Although, preheating the board is much more important, since this will reduce the soldering time considerably, reducing the thermal shock as well. Might as well put the MLCC's sitting somewhere on the board to preheat as well, "for free" - that won't hurt.
Preheating the board to about 100degC for manual assembly makes a lot of sense, since this is a temperature the FR4 board and already placed components (such as electrolytic caps) can take basically forever, with a good margin, so you don't need to think about any negative side effects.