All the BNC sockets I have here are 75 ohm. But if the supplier had inadvertently shipped 50 ohm on a production batch it would not be a problem.
This is because the errors from impedance mismatch depends upon how long the mismatch is against the rise time of the signal. So a 50 ohm connector would be a chain of 75 ohm source, then 50 ohm connector, then back to 75 ohms. The 75>50 ohm transition would create a reflection, followed by an inverse (equal and opposite) reflection from the 50>75 ohms, cancelling the first reflection. The separation on these reflections would be about 100pS - but this is tiny compared to the signal rise time of about 2000pS. This would mean that the glitches would be negligible in value and of a small enough time duration that the DAC could not possibly respond to it - too small in voltage and way too fast for the receivers to respond.
So it's a complete non issue. However, the cables should be 75 ohms, as a 1m length has a delay of the order of 5000pS, and this would create voltage errors that the receiver would respond too, as the glitch levels would be higher than the hysteresis thresholds on the receiver and the time period would also be long enough for the receiver to respond. That said, I have built in de-glitch circuitry in the SPDIF receiver to accommodate these kind of errors.