Why did a repair that held for months suddenly break? (fatigue, not bad glue)

Evidence grade: A2 — Institutional literature Effort: N/A — reference, no procedure

It didn't fail because the glue was bad — it failed because the load was repeated. Adhesives can be very strong in a single pull and weak under thousands of small cycles, and no static-strength number predicts cyclic life.

Static strength is not endurance

This is the single most misunderstood fact in adhesive repair. A structural adhesive's data-sheet strength — the lap-shear number everyone quotes — is measured in one slow pull to destruction. Real service is different: vibration, flexing, thermal cycling, a door slammed a thousand times. Under repeated loading, adhesives accumulate damage microscopically and can fail "prematurely" — long before their static strength would suggest (source 1).

The industry's own testing proves the disconnect. Henkel's fatigue study of structural adhesives (2019) tested multiple chemistries under cyclic loading per ASTM D1002 and D3166 and found that high initial lap-shear strength did not predict fatigue life — the tested chemistries had materially different endurance, and in that specific test set, the high-performance epoxy retained much less of its initial strength at endurance than several alternatives (source 1). Read that carefully: the strongest glue in the one-pull test was not the longest-lived under cycling.

How fatigue actually kills a joint

• Peel is the enemy. Adhesives are strong in shear (pulled sideways across the bond) and weak in peel (pried apart at an edge). Cyclic loading almost always introduces a peel component — a corner lifts a fraction of a millimeter, the crack starts at the edge, and each cycle walks it forward.
• Vibration is fatigue. A bracket on a running engine doesn't experience "a load"; it experiences millions of tiny load cycles. What looks like a static joint to you is a fatigue specimen to the material.
• Geometry decides. Long overlaps, tapered ends, and joints loaded in pure shear live far longer than short overlaps with square ends loaded in peel. The same adhesive in two different joint designs can have wildly different lives.
• Stress concentrations start the clock. A void, a sharp corner, a bond-line edge — fatigue cracks start where stress concentrates, then propagate. This is why prep and joint design matter more than brand.

What this means for your repair

1. Never promise a lifespan for a purely glued joint under cyclic load. If it vibrates, flexes, or cycles thermally, the honest answer is "I don't know how long" — anyone who gives you a number is guessing.
2. Design for redundancy. The professional answer to fatigue is mechanical backup: rivets, screws, or bolts alongside the adhesive, so the joint doesn't depend on one failure mode.
3. Eliminate peel in the design. Clamp the joint so loads go through it in shear; avoid cantilevered tabs glued at one end.
4. Suspect fatigue, not the adhesive, when a good repair dies months later. If the prep was right and the cure was right, and it held for months before letting go suddenly — that is the signature of fatigue, not a bad batch.

Honest limits

• The Henkel results are product- and joint-specific — they prove that static strength doesn't predict fatigue life, not that "all epoxies are poor in fatigue." Different chemistries rank differently, and your joint isn't their test coupon.
• Real fatigue data for consumer repair adhesives is essentially unpublished. Manufacturers publish static numbers because they're easy; endurance data is rare, expensive, and joint-dependent. Treat any lifespan claim with skepticism proportional to its precision.
• This article explains the failure mode; it is not a design guide. Sizing a fatigue-critical joint needs engineering, not rules of thumb.

Sources

1. Henkel — "Fatigue Resistance of Structural Adhesives" (2019) — https://www.hisco.com/userfiles/images/landingpages/henkel/structural-adhesives/Fatigue-Resistance-of-Henkel-Structural-Adhesives.pdf — fatigue under repeated/dynamic loading causes premature failure; high initial lap-shear strength does not directly predict fatigue life; tests per ASTM D1002 and D3166; tested chemistries had materially different endurance; in that test set the high-performance epoxy retained much less of initial strength at endurance than several alternatives; results are product- and joint-specific.

Repair Lab · Release 1 · Evidence last checked September 30, 2026 · A project of AOJEL