A divergence can exist for months, renewing at every new extreme, while the trend carries calmly on. That is neither a markup failure nor a "false" mismatch — it is normal behaviour, and it follows from how a mismatch is built.
Why it renews
A mismatch is built on two neighbouring extremes. A new peak appears, the comparison is made against it, and if the indicator again reads lower, the mismatch arises afresh.
In a strong trend that is exactly what happens: price prints peak after peak while the indicator answers more weakly each time. Formally it is a chain of separate mismatches; in substance one long observation — the move continues while the measured impulse falls.
Why that is not a contradiction
Because a divergence says not "the move will end" but "it is running at a lower speed than before". Those are compatible: a decelerating move can keep moving for a very long time.
Oscillators are bounded above and below by construction — they cannot climb indefinitely alongside price. In a prolonged trend the indicator runs into its limit, and after that almost any new price high is bound to produce a mismatch. It arises not because the market weakened but because the measure ran out.
What that says about the measure
The main point: a mismatch is a property of the pair "price and a particular indicator", not a property of the market.
Hence a practical consequence. A long-standing mismatch is better read not as accumulating evidence for a reversal but as a sign that the chosen measure has stopped distinguishing situations: if it gives the same answer at ten peaks in a row, that answer carries little information.
Where this is seen most
On prolonged trends without deep pullbacks — there mismatches come one after another. And on instruments that travelled many times over: price grew by orders of magnitude while the oscillator lived in its narrow band throughout.
How to handle it
Look not at whether a mismatch exists but at what changed besides it. A divergence in its fifth month and one appearing for the first time after a long stretch of agreement are different observations, even though by definition they are identical.
The second useful move is to check what another measure says. If one indicator sees a mismatch and another does not, the conclusion depends on the choice of indicator rather than on the market.
A common mistake
Calling such a mismatch false and concluding that divergences "do not work". They were never asked to: they describe the past. What was false was the expectation, not the observation.