I've been testing various approaches, and find that, if you have an adequate Vcc supply, the incidence of problems, particularly with oscillator startup, is much less frequent.

As for the resistor, the datasheet for the particular MCU you're using will tell you whether you need it.

If your Vcc-to-Gnd capacitance is small (<100 uF) and your Vcc rise time is less than 10 ms, the RC approach seems to work OK.

Presently, I've reached nearly 10 million iterations, yet haven't had even one external BBRAM corruption problem, though I used to have one nearly three times in 10 with a smaller power supply, when I've used a mosfet to switch on the Vcc and another mosfet to discharge it to GND each time the reset switch was closed or the raw supply is decaying, and thereby ensure that the fall time of Vcc is on the order of 4 microseconds. My results are very preliminary, however. My current experiments are intended to determine whether these efforts lead to any predictable result

As time allows, I'll have better, and possibly more precise, information once I set up a formal test. Simpler circuitry is desirable, but I think, so far, at least, that the capacitance on Vcc is the critical parameter. If it is too large, the decay time of Vcc is too long, and the MCU occasionally runs away despite the assertion of RESET.

RE