Is a standard PLL required when a step down transformer produces a in-phase waveform of the grid voltage?
A first reason is that, a step down transformers has a lot of losses and distorts the waveform as the magnetizing current is not linear, I expect an almost triangular voltage with a small transformer. One could use another type of galvanic separation or a differential amplifier. But the next problem is even worse:
"Voltage proportional to current" creates a negative resistor in grid injection. This will oscillate at some higher frequency where the grid Thevenin parallel resistance is higher than the V/I rating of the converter.
For a PFC power factor correction circuit as a load it is even good: it will damp harmonics, buy for a grid tied inverter: no way.
Dependent of the inverter power, if your inverter is of very low power (ej: 100w-500w), surely, you must use easy algorithms and low software resources, where the transformer reduction is a aceptable option for the operation.
In other hand, the greater the power of your inverter, you will must require more resources software, measurement and more sophisticated algorithms, where the PLL is required.
Many small power devices is the same problem as one big power one. A PLL is not required as such, but negative differential resistors make the grid unstable. Using a zero crossing after a low pass or band pass filter are possible as well, but not "voltage proportional to current" as it results in a negative resistor. Two small converters in parallel with a negative resistance and a high impedance at some frequency in the grid (resonance) is enough to get the grid oscillating.