if you want to reliably drive a latching relay, use a microcontroller.
Here is one such example.
PedalPCB has a
latching relay bypass board that uses an NE555 (instead of a microcontroller) to maintain state and react to momentary switch presses. I think the NE555 could be replaced with a couple inverters (e.g. cd40106 or 74hc14) with a
circuit like this. I have breadboarded this, and it certainly works for toggling between two states with a momentary switch. The advantage is dramatically reduced power consumption (the disadvantage is a physically larger chip).
With the single-coil latching relay, you need the ability to produce a bi-directional current pulses (pulse one way for set, pulse the other for reset). That is easily done with two pins of a microcontroller (that have the appropriate current sourcing/sinking capability and builtin protection from coil flyback). The way it's done with the PedalPCB bypass is to use a capacitor in
series with the relay coil. While DC won't pass through the capacitor, there will be a current pulse generated by it being charged or discharged depending on which transistor is open (either the PNP to ground or the NPN to VCC).
My question: assume that Q2's collector is attached to the relay coil's design voltage (5v in the case of the TQ2-L-5V, which also happens to be a good supply voltage for inverter ICs). Could we then remove the series capacitor C3, and replace R4 with a series capacitor (i.e. in series with the base of Q1)? And additionally add a resistor to ground between transistor gate and the series capacitor. The goals being (1) to activate the relay coil with it's design voltage, instead of the approximately-correct voltage from the cap charge/discharge current, and (2) get away with a smaller capacitor (hopefully small enough to use film or ceramic) since we only need to drive the transistor base?
I have bread-boarded something conceptually similar, though I'm using a
two coil latching relay, and MOSFETs instead of BJTs - but the main idea from my question is the same: using a series capacitors to generate a pulse for the FET gates (and the FETS do the "heavy lifting" of current sinking). It works as expected on the breadboard. Assuming this isn't a flawed design for some reasons that I'm unaware of, for simple on/off relay bypass, it has a lot of advantages: micro-amp current consumption (except when switching of course), simple circuit, uses cheap, commodity, current-production components. Schematic attached.