To put slightly more detail on it: it's the gold that's not dissolved, that's a problem. Free dissolved gold, in solder, doesn't do anything. If it can't dissolve, then it will form brittle intermetallics, that are a problem.
Gold may not dissolve, for two reasons:
1. Too quick reflow cycle. The joint ends up nonuniform. Basically, there wasn't enough solder, time and temperature, to get it dissolved. Probably, one side of the joint has partially dissolved plating or intermetallics, forming a stress raiser, leading to peeling failure or solder cracking.
2. Too much gold, solder becomes saturated. Gold remains, and intermetallics form on the plating boundary, leading to peeling failure; or intermetallics precipitate on cooling, leading eventually to cracking of solder joint/ball.
Example, BGA is better than LGA, because the balls can dissolve a lot of gold. They need to be kept molten long enough to succeed. LGA might not be able to dissolve that much. LGA, and QFN, SON, etc., all use the same type of joint.
Repeat the logic for nickel or other barrier metals. AFAIK, nickel dissolves very little, and very slowly, making it an acceptable barrier. It would be used alone, except it's nearly unsolderable due to its surface oxide.
In contrast, copper and silver simply dissolve in solder, and precipitates (on freezing) as an innocuous intermetallic (tending to enhance strength rather than brittleness). Tin plate, silver plate, HASL and OSP are all attractive finishes for this reason. HASL may be undesirable for flatness reasons (important with large LGAs, QFNs, etc.), but the others should be good.
Tin plate, silver plate and OSP have limited shelf lives, but as you said the assembly will be prompt, this is not an issue, so these are acceptable.
You may ask your PCB fab about what they recommend, or commonly produce, for military and space applications. Needless to say, if you're asking a fab that's not ITAR qualified, you're probably not talking to the right people.
Tim