The latching relay closes by a pulse in one direction, and resets by a pulse in the other direction. So the coil may generate a spike in either direction. We mistakenly might try to solve this by putting two anti-parallel freewheeling diodes across it, but then they rob current from going through the relay coil.
It could be worth a try to install two back-to-back zener diodes, whose rating is slightly greater than the supply voltage. However it's unknown: a) how effectively zener diodes reduce spikes, b) how well they endure repeated spikes, and c) how well they protect neighboring components from over-voltage bursts.
Ordinary silicon diodes are robust and drop the voltage to 0.6 V.
I think the schematics depict different approaches using plain diodes, to solve the spike problem.