What do you think of this guys idea of adding a tank to the tankless system?
http://www.finehomebuilding.com/PDF/Free/021192082.pdf
A tiny buffering scheme like that fixes the cold water sandwich and flow issues (some of the tankless heat exchangers have ridiculous head load at 8gpm.) Seems like an expensive solution for those problems. This is the configuration I'd suggested NOT using with the buffer located at the remote sinks due to the high distribution losses of a very long pump loop (twice the plumbing, twice the distribution loss, more or less.) It works though- and done right it eliminates short cycling.
In order to get the full rated heat out of it you need to really put a monster pump on it or crank up the tankless output temp though. I'm not sure if the hysteresis in the aquastat of a typical electric mini designed to control a 1.5kw element is going to provide suffiently long burns to not short cycle at the full burner rating- there'll have to be some compromising of flat-out capacity to tweak the overall efficiency. It'll modulate some at high flow, but not a lot. Rather than using a power hog like a Taco 09 they used I'd go with something smaller (maybe even a multi-speed to be able to tweak it) and raise the output temp of the tankless to get the required BTU/hour rate into the tank. That would potentially give you a bigger modulation range out of the tankless too, allowing it to better match the instantaneous load. (Some tankless heaters are less flexible than others in their output temp selections, but they've gotten better over time.)
It's not different from a small indirect tank and a boiler other than in this instance the "boiler" is the tankless, and it's heating recirculating potable water directly (not through a heat exchanger.) At some point, between tankless loop pumps and recirculation loop pumps, buffer tanks & tempering valves etc you have to just get a grip, and ask yourself why you're making it such a g'dawful kludgy Rube Goldberg contraption just for hot water. It's one thing if it were incorporated into the heating system to improve the efficiency of both heat & hot water, but IMHO it's a bit silly (and expensive up front) for just the hot water end of it.
Some newer "tankless" on-demand heaters now incorporate tiny ~1 gallon header-buffers to eliminate the cold water sandwiches and provide a minimum burn length to increase the net efficiency (no ultra-lossy short cycling), at the expense of a small amount of standby loss. A tiny buffer like that would reduce it's efficiency a small amount in an EF test due to standby losses, but would increase it's efficiency significantly in real-world use patterns. The current EF test draws are all 10.2 gallons at a time- that's exactly how YOU use hot water, right?
The very real very significant short-cycle losses of tankless heaters never show on the EF test, a bone of contention with tank heater manufacturers. (The test standards are rightly undergoing some revision...)