The constructor for an SPI object always calls init() supplying all arguments but the SPI id. So with respect to the SPI object there should be no difference. But there is a subtle side effect between using a single call and splitting set-up in two calls. Even with the call machine.SPI(1) the pins for mosi, miso and sck are set, as well as the baudrate, but in that case the default values are used, which are baudrate=500000, polarity=0, phase=0, bits=8, firstbit=0, sck=14, mosi=13, miso=12 for a Generic ESP32. Maybe this causes the difference.

other than to see that the 'constructor with params' method does not merely construct then call init so errors are bound to creep in

Both machine_hw_spi_init() and machine_hw_spi_make_new() call machine_hw_spi_init_internal() after a formal check of the argument list.