This is an example of what happens when I have two somewhat unrelated things in one sentence and then forget to edit properly before posting...
Rather obviously the amount of heat transferred from exhaust steam (at something like 14-18psi) is not going to be anywhere near saturation at boiler pressure. As Prof. Milenkovic notes, for this you need a combustion-gas heater (aka 'economizer') which, in turn, requires some form of effective pumping to pressurize the feedwater to keep it from 'steaming' (look up shipboard 'steaming economizers' for a case where that phenomenon might be desirable, and consider the reasons why it would not be on a locomotive). That system might be a positive-displacement pump or the kind of pitot pump used for some "BFP" applications on once-through/supercritical powerplant boiler setups.
To get close to the atmospheric boiling point of 212 F let alone the water/steam saturation temperature that you quote of 339 F at boiler pressure, you probably need some 2-stage scheme, possibly even compound expansion where you divert some of the steam from the intermediate pressure receiver into a feedwater heating stage.
With the aforementioned point that there is some effective 'superheat' in exhaust steam at 'back pressure' relative to 212F, which allows a somewhat higher temperature from the hot-water pump into the boiler, yes, you'd need bleeds. However, I doubt there is much to be gained at 'typical' locomotive compounding pressure, contrary to the situation in a good powerplant heat-balance arrangement with bleeds taken directly from turbine interstages to implement both steam reheat and feedwater heat.
Even in a compound arrangement, expect boiler maintenance to become progressively prohibitive as you get much above 310-315psi; I think this would be true even if 'fillet-welded staybolt' installations turn out to work as desired in large North American fireboxes. Here the interstage (receiver) pressure is not likely to be terrifically high: in fact, I suspect just the opposite procedure is likely to be used (the injection of boiler steam, at working-pressure saturation temp and hence effective superheat, into the receiver proportionally so that the working thrust for the LP balances that from the HP correctly).
A feedwater heater is helpful in not only reducing fuel consumption but also reducing the "water rate" by 7-10 percent. This may reduce fuel consumption by even more because you are placing less demand on the firing rate to the grate and hence may get less carbon carryover up the stack.
Note that you need to arrange in a closed FWH to recompress the heat-exchanger condensate to get the water-rate gains. I think it is much easier to use an open setup where the hot-water pump only has to deal with one 'feed source' that runs at whatever pressure the mix of exhaust steam and tender water develops within the heater, although the necessary venting of air creates a certain unavoidable loss of steam/water too.
The point about reducing firing to a level well short of the grate limit is a good one, related to the principle behind using a large, wide firebox (with full circulators) and all the extra carried weight that construction implies -- it is the primary reason for those six-wheel trailing trucks in Lima designs -- and I want to encourage recognition that the less carryover of any kind is wasted up the stack, or in premature combustion quench producing sooting, the better.
There was that Franco-Crosti scheme tried in England and other places that was "Rube Goldberg" and had the added problem that it took too much heat from the flue gases that you got acidic condensation and wrecked equipment.
Not as Mickey Mouse as it seemed (and part of the answer, if you find it useful to get the thermo gains, is to use dolomite or a similar sulfur-fixing additive in the fuel or make components in the 'cold end' of the gas path out of noncorroding material). Note that there are advantages in Rankine-cycle recovery of gas-plume heat well below the effective dew point of 'sulfuric' acid compounds. There's also great advantage (getting a bit ahead here) in incorporating some of the air preheat into the same shell used for countercurrent economizer heating; this may work well for some of the secondary air requirements.
A much better scheme was devised by Chapelon, I believe, and considered by Wardale for the unfunded 5AT mainline "tourist train" locomotive scheme. You just put a barrier inside the boiler water space to reduce the mixing between the front part of the boiler where the flue gases have cooled a little and the back part of the boiler, and you pump your feedwater into the front part. That allows the downstream flue gases to give up more of their heat and cool closer to the atmospheric boiling point by heating feedwater below the boiler saturation temperature.
This is one of those Porta things (IIRC he called it part of his 'sectional boiler' approach) that looks really good until you start figuring the differential stress on parts of the boiler and tube/flue system. Joe Burgard did some good preliminary calculation on this setup for the T1 Trust's locomotive, and the 'leaky partition' winds up being only a couple of feet from the front tubeplate. When you factor in gas speed and heat transfer at that point in the flow, across that distance, the actual heat drop in the gas and hence the amount of incremental feedwater heating from the arrangement may not be "that" substantial. What it gives you more of is a controlled way to introduce relatively high feedwater flow (for high steam-generation rate) without shocking the boiler or creating weird internal flow changes in convection-section circulation.
You don't want to cool the flue gases below the atmospheric boiling point because then you are back to the maintenance problems of the Franco-Crosti system.
Let me repeat that the critical gas temperature is nowhere near as low as the atmospheric boiling point of water. It's the sulfuric acid condensation point, perhaps complicated by presence of SO3, that is the concern. The problem for any Franco-Crosti heater of 'conventional' design is that different firing levels result in the gas plume reaching this temperature at different points within the heater shell; if the heater is designed 'never' to reach that point even in idling operation, anything close to full output will be throwing away enthalpy at heroic rates...
Oh, and you probably want a combustion air preheater while you are at it, using heat from exhaust steam just like a feedwater heater and also condensing some of that diverted exhaust steam and achieving additional water rate savings...
Actually, what you may do 'best' with is a Snyder preheater, which is essentially a closed heater made out of loops of plain pipe suspended in the primary-air spaces between the ashpan and the water legs. Exhaust steam here (which is what Snyder proposed to use) should produce air temperature very close to saturation temp at effective back-pressure in the tube circuit at normal primary air flow rates, with little effective need for moving parts. The setup was tested on C&O in the late Forties, with what I remember as upward of 10% efficiency gain (one thing it does is heat the 'inert' nitrogen component of the primary air without drawing any of the combustion heat from the firebed, a decidedly nontrivial effect).
... if you can get the intake air closer to 212 F from the bottom end, and if you can get spent flue gas closer to 212 F from the top end in the economizer, you stand to approach 100% boiler heat transfer efficiency (at least with respect to the "low heat value" of the fuel because gosh no, you don't want to start condensing flue gases in a steam engine even though you do that in your home high-efficiency furnace.
The 'correct' approach to intake-air preheat (apart from Snyders in the primary airflow), in my opinion, is to use the economizing principle together with direct FGR; this is something that can be facilitated by some of the 'usual' layouts for Franco-Crosti equipment. It's a little like an 'open FWH' for the combustion air, and the recycled 'spent gas' has the effects noted in powerplant operation, but without the large overhead in weight and packaging volume involved with most FGR arrangements there.
In theory, "closed shell" heat exchangers used for either feedwater heating or combustion preheating (never implemented on locomotives but talked about) could go higher than 212 F. Using check valves, you could receive the "blow down" steam just as the exhaust valve opens at somewhat higher than atmospheric temperature, especially when the engine is working hard at long cutoffs and wide open throttle. But owing the required temperature differentials in practical heat exchangers, exceeding 212 F deriving heat from cylinder exhaust in a simple-expansion engine is a difficult proposition.[/quote]
Perhaps a better source -- one that I hadn't really considered -- might be to take the exhaust from actual boiler blowdown, including continuous blowdown, and perform the final stage of feedwater heating, just before admission to the boiler through check valves, with countercurrent flow.
I do not think you can effectively capture the momentary high pressure of compression in most piston-valve setups ... at least, not cost-effectively for the additional level of thermodynamic 'savings' available. Might be different for a proper arrangement of poppet valves, or if Carter's reversible compression storage arrangement is in operation (you would bleed off the 'high-pressure' steam selectively rather than modulate it back into the cylinder and tract). I tend to look at compression optimization for smoother running at high speed/high cyclic, so I haven't tinkered with the idea for a fiddly additional bit of thermodynamic enhancement.