I'm not one to suggest that the RC reset is the be-all and end-all for 805x resets, but I'm not persuaded that supervisors of any sort deal with 805x reset adequately, as they really only deal with power-on reset, which RC does just fine. The only aspect in which they improve on the RC is in that they filter out the problems caused by ripple on the often-dodgy power supply.

I've got a couple of examples with which I'll play as time allows, but the 1/2-amp "simple switchers" certainly don't "cut the mustard" from what I've seen so far, and wall-warts are widely variable in their usefulness. Startup is often the worst-case for the PSU. Too slow a rise time can lead to a dodgy RESET as well as trouble getting the oscillator to start.

The absence of a reliable reset, which is an "attribute" I've found in several Intel products since they were first released, explains why most of my MCU implementations over the past three decades have contained only non-Intel products. The ones that did have Intel parts, had parts that use negative-going reset. I found that positive-going reset, interrupts, and the like were silly, since they couldn't be wired-ANDed and since TTL really couldn't drive an active low to a high. BTW, the RC-reset does work OK with bidirectional resets, as some 805x products have. Dealing with a bidirectional RESET for under $1 is a real challenge if you want a reset circuit that works properly under Vcc decay and Vcc brownout conditions.

What does your MCU do when Vcc falls below its Vcc minimum? Is that guaranteed by the manufacturer? Have you ever observed the signal MCU behavior during Vcc decay? Have you ever observed signals that you didn't expect?

If Vcc is out of spec, what is the MCU going to do with a RESET that's not at a valid Vcc level? How will it interpret the signals from other internal resources, flash, etc?

Since the RESET level is dependent on Vcc, I suspect you can't get a valid RESET when you don't have a valid Vcc. I'm of the opinion that RC is an adequate power-on reset. I don't believe it deals with slow decay or momentary brownout on Vcc. For that reason, among others, I believe the correct and truly safe way to deal with brownout and Vcc decay on power-down, is to ensure that the rise time of Vcc at power-up is very short, meaning less than 1 ms, and that the decay of Vcc at power down is short, meaning less than 10 us. Even short fall time might be smarter. I believe, further, that this can be implemented for a small cost, probably not as small as that of a low-cost "tripod" as Erik has suggested, but it deals with the fall of Vcc as well as the rise, and it ensures that the MCU won't do off doing something harmful during a slow decay of Vcc, yet it tolerates the large-capacitance, low-current-power-supply situation that's common these days.

None of my old Intel MCU applications use FLASH. They predate that by at least a decade, though I've plunked FLASH-based 805x drop-ins in their circuits with little trouble. The majority of those were either early microprocessor boards (8080), or 8048-family members, all of which used negative going reset, hence, had no such problems as do the 805x's. I'm a big believer in single-point grounds and in ground-referenced controls.

I believe it might be worth a couple of flip-flops, an extra 5-volt regulator, a few passives, and a couple of mosfets to attain a truly reliable reset for an 805x, though the cost would have to go up for a part with bidirectional RESET. One of the popular supervisors might even have a useful role in such a fix.

RE