Why a seal fails when it is too tight
Over-compression, and why more squeeze is not more sealing.
A seal fails when it is too tight because an O-ring squeezed well beyond its intended amount stops behaving like an elastic material pressing evenly against its surroundings and starts behaving like an overstretched one, thinning, losing its ability to spring back into full contact after the smallest disturbance, and in some cases tearing or extruding into whatever gap happens to be nearby, all outcomes that leave the seal worse off than a properly, moderately compressed ring ever would be. This sits directly against the instinct the previous two articles have already built up, that a seal is entirely a story of squeeze being applied correctly, and it is worth stating plainly here that correctly cuts both ways, too little and too much are both real, distinct ways for the same joint to fail.
Introduction and overview
Squeeze works by keeping a ring's rubber under constant elastic strain, pressing back against the surfaces around it because it genuinely wants to return to its original, larger cross-section, and that elastic push depends on the rubber staying within its comfortable working range of deformation, the same range that lets it recover fully and repeatedly rather than being pushed toward its actual limits. Over-compress the ring and it is forced well past that comfortable range, thinning out along its cross-section in a way that reduces rather than increases the contact pressure it can actually maintain, and in more severe cases the rubber is squeezed hard enough that it starts to flow, under pressure, into any tiny clearance gap available at the edge of the groove, a failure called extrusion that can slowly nibble material away from the ring with every pressure cycle until a groove that was correctly sized on day one is effectively starved of ring material months later. It is a genuinely counter-intuitive result for anyone used to thinking of a tighter fastening as a safer one, since almost everywhere else in mechanical assembly, a bolt torqued harder or a clamp closed tighter reads as more secure, while here the same instinct, squeezing harder for extra confidence, actively works against the seal it was meant to protect.
The memory-foam comparison
A memory foam pillow left compressed for a long time under a heavy stack of books does not spring straight back to its full original loft the instant the books are lifted off, it recovers slowly and, if it was compressed hard enough for long enough, incompletely, settling into a slightly flatter resting shape than it started with even though nothing about the foam was ever cut or torn. An over-squeezed O-ring is subject to exactly the same kind of permanent, cumulative damage, called compression set, in which rubber held too tightly for too long gradually loses some of its ability to spring back to its original shape once the compression is relieved, which means an O-ring that has spent a long service life over-compressed can end up sealing worse, not better, than a ring compressed correctly from the start. A pillow given regular breaks from the weight stacked on it recovers its loft far better over time than one left permanently loaded, which is the same reasoning behind cycling a seal's compression down deliberately during long periods of storage where possible, rather than assuming a rubber ring left compressed indefinitely will simply wait patiently, unharmed, until it is finally needed.
The one number worth remembering
The comfortable squeeze range described in the first article of this set, roughly a fifth to a third of the ring's own cross-sectional thickness, has an upper edge for a real reason, since compression beyond that upper edge measurably increases the risk of extrusion and accelerates compression set, and the difference between a squeeze at the safe upper end of that range and one meaningfully beyond it is often smaller, in absolute terms, than the width of the tolerance band a careless groove dimension could easily fall outside of, which is precisely why the groove tolerances the previous article described are held so tightly, the safe window they are protecting is genuinely narrow on both sides at once.
What this changes in practice
Because more squeeze is not simply a safety margin in reserve, gland dimensions are specified with both a minimum and a maximum in mind rather than only a minimum, since a machinist or designer thinking purely in terms of avoiding an under-sealed joint might reasonably assume that a slightly deeper, tighter groove is the safer direction to err in, when in fact erring too far in that direction introduces a different failure mode entirely. Anyone reviewing a gland drawing has to check both bounds, not simply confirm that the groove is deep enough, since deep enough and too deep are separated by a genuinely narrow, easily crossed margin. That habit of checking both directions, rather than assuming more of a good thing is automatically safer, is worth carrying well beyond seal design, since a surprising number of engineering failures trace back to exactly this same one-directional assumption applied somewhere it did not actually hold. Torque on a bolt is the most familiar cousin of the same trap, tighter feels safer right up until the fastener is stretched past its own elastic limit, and a designer who has already internalised that lesson for bolts is still, surprisingly often, caught out applying the opposite, more-is-always-better instinct to a seal instead.
Where this stops being true
The risk of over-compression depends heavily on the specific rubber compound used, since some materials tolerate a wider range of squeeze without significant compression set or extrusion risk than others do, and a material chosen specifically for its resistance to those failure modes can afford a squeeze range a more ordinary compound could not safely match. That material-specific tolerance is exactly why gland dimensions are specified alongside a named material rather than treated as a universal geometry that works identically for any rubber dropped into it, a groove correct for one compound is not automatically correct for another, even at the exact same nominal squeeze, which is one more reason a gland drawing has to name its intended material explicitly rather than leaving it as an assumption anyone downstream might quietly get wrong.