Helium-3 is the abundance that refuses to evolve, and its refusal indicts the models that must evolve it. BBN produces 3He/H near 1.0 x 10^-5, and galactic chemical evolution should then push it upward over cosmic time, since low-mass stars are calculated to manufacture 3He copiously and return it to the interstellar medium. The measurement, from the 3.46 cm hyperfine line of ionized helium-3 in Galactic HII regions (Bania, Rood and Balser 2002, and successors), finds 3He/H flat at about 1.1 x 10^-5 across the Galactic disk, independent of metallicity and galactocentric radius: no enrichment, anywhere, over ten billion years of stellar processing. This is the "3He problem": stellar yield calculations and the abundance record cannot both be right.
The standard reconciliation invokes extra mixing in red giants, thermohaline convection destroying 3He in the very stars that should produce it, calibrated so that net production approximately cancels. The cancellation must be fine: too little destruction and the disk over-enriches, too much and BBN's relic is eaten. That a nucleus survives at its primordial value purely by accident of opposing stellar processes is the kind of balance that earns suspicion, and the mixing physics, while supported by carbon isotope ratios, remains parameterized rather than derived. Meanwhile the flatness itself is the saving grace for cosmology: because 3He neither grows nor shrinks, its local value still reflects the primordial one, confirming the BBN network at the ten-percent level.
The standing is a quiet decades-old anomaly of stellar physics with cosmological stakes: the balance holds in every direction measured, the calibrated mixing remains the only account, and resolved observations of 3He in planetary nebulae keep returning yields that deepen rather than relieve the contradiction.