Two pairs of headphones on one output. The forum answer is “you lose half the power”, and that is wrong — with a perfect amplifier nothing changes at all. Both pairs see the same voltage they would have seen alone, and both get the same power. Every bit of loss comes from the amplifier's output impedance not being zero.
The real problem is usually somewhere else entirely, and this shows you where.
The low-impedance pair wins. Two headphones in parallel present the amplifier with less than either alone, and the lower one dominates that figure. Put a 16Ω in-ear alongside 300Ω over-ears on a source with 10Ω of output impedance and the over-ears lose over 4 dB while the in-ears barely notice. The pair you were trying to share with gets the worse deal.
Sensitivity is the real killer. Level is set by voltage, and voltage sensitivity is what matters when both pairs see the same voltage. A typical in-ear can be twenty-five decibels louder than a pair of 300Ω over-ears at the same setting — there is no volume position at which both are comfortable. That, not impedance, is why sharing usually fails.
Current adds up. The amplifier supplies the sum of both currents, not the average. A portable source that was comfortable with one pair can hit its limit and clip with two. This calculator models an ideal voltage source, so it will not show that happening — compare the total current it reports against what your source can actually supply, which the power curve works out.
Never split a balanced output. On a balanced or bridged jack both conductors are driven — there is no ground. An ordinary splitter shorts them together and can destroy the amplifier. This applies to 4.4 mm and 2.5 mm outputs, and to any adapter that ties one leg to ground.
The argument above is arithmetic, so it can be checked against the 567 models and the 169 sources here that publish an output impedance. Take the hardest case anyone actually runs into: a 16 Ω in-ear sharing a socket with a 300 Ω pair of over-ears. Across those 169 sources the median loss on the over-ears is 0.37 dB — inaudible, and nothing like half the power. 38 of them lose more than a decibel and 34 lose more than three, which is the real shape of the problem: it is rare, and where it happens it is severe. The worst in this database is the Bottlehead Crack (OTL kit) at 120 Ω, where the over-ears drop 16.1 dB the moment the in-ear goes in.
The sensitivity problem is the one that bites every time, and it is bigger than the figures above. Measured per volt — which is what matters when both pairs see the same voltage — the median in-ear here is 122 dB/V against 112 for the median full-size model, a gap of 9.9 dB. Across the whole database the spread is 46 dB, from the Simgot EA500 at 141 dB/V down to the HiFiMAN Susvara at 95. No volume control reconciles that; a splitter with its own attenuation on one leg, or two amplifiers, is the only thing that does.