thermal-expansion mismatch
Glue two strips that expand at different rates, heat them, and the assembly bows — the fast-expanding side gets longer and has to curve around the slow side. When two bonded materials disagree about how much to grow with heat, the disagreement shows up as stress right at their shared boundary. Sometimes that is useful (a thermostat), but far more often it is a failure waiting to happen.
The mismatch strain is roughly (alpha_1 - alpha_2) x delta_T, and it drives a shear and peel stress at the interface. A silicon chip (alpha about 3 x 10^-6 per K) soldered to a copper board (alpha about 17 x 10^-6 per K) and cycled 100 degrees C sees a mismatch strain of about (17 - 3) x 10^-6 x 100, which is 1.4 x 10^-3. Every thermal cycle works this back and forth — the classic cause of solder-joint fatigue in electronics. The bigger the difference in alpha, the bigger the delta_T, and the stiffer the materials, the worse it gets.
The whole art of joining unlike materials is managing this mismatch: match the coefficients (special glass-to-metal sealing alloys like Kovar are tuned to match borosilicate glass and silicon), insert a compliant interlayer that absorbs the strain, grade the composition gradually, or keep the parts small so the absolute movement stays tiny. Mismatch governs the reliability of electronic packages, glass-to-metal seals, dental fillings, coatings, and every joint between a ceramic and a metal.
Glass-to-metal seals in vacuum tubes and light bulbs use matched pairs (soft glass with Dumet wire, or borosilicate with Kovar) precisely so the two shrink together on cooling; a mismatched pair would crack the seal and leak.
Reliable seals and packages come from matching expansion, not from making a joint stronger.
Making the joint stronger does not fix mismatch — it can make it worse, because a stiffer bond transmits the mismatch strain as higher stress. The cure is to reduce delta_alpha, add compliance, or shrink the part.