minor sub-fields

In my recent semi-rant about citations, I noted that one issue is that papers in small sub-fields are intrinsically incapable of racking up large number of citations (since not that many papers about that subject are published), unless the paper becomes interesting and relevant to a larger sub-field.
This incidentally is why high citation biology papers tend to have order of magnitude more citations than highly cited physics papers. For a physics paper to get 1000+ citations is extraordinary, whereas top biology papers get over 10,000 citations
(see eg ScienceWatch). In fact there are no physicists among the top 50 cited scientists, although this is in part due to biologists ability to write more than 1000 papers each...

So what? Why not squeeze out the small sub-fields? Clearly they don't matter to most people, and the few that do stuff relevant to the bigger community can be duly rewarded.
Well, no. Don't work that way. And in fact I'd argue this is a trap that the greater flexibility and responsiveness of the US science community is vulnerable to - the more stodgy, stay-the-course European style of science tends to let them ride out faddishness and sustain broader ranges of sub-fields.
Here is why that matters:

first - little fields can become big fields with very short notice; in physics, a trivial instance is the search for high temperature superconductors, or Bose-Einstein condensates; in astronomy, the most recent instance is the theory and observational search for extrasolar planets.
How does that work - well, consider for example the well known Goldreich & Tremaine 1980 ApJ v241 p425 paper on disk satellite interactions. A paper now recognised as predicting the essential physical mechanism of planet migration through coupling to the gaseous disks.
ADS shows 277 current citations (an undercount, but representative). 225 of those citations came in the 10 years since the discoveyr of the planet around 51 Peg in 1995 - only 52 citations came in the first 15 years the paper was published!

Similarly, the basic technique used by Mayor & Queloz 1995 (702 ADS citations) came from a sequence of work over 20+ years; including Marcy & Butler 1992 (78 ADS citations), Cochran & Hatzes 1991 (10 ADS citations), Campbell & Walker 1979 (51 ADS citations) and Griffin 1973 (61 ADS citations).
The point is that incremental advances in a relatively minor field over 20 years lead to a breakthrough which is now one of the largest observational and theoretical sub-fields in astronomy, and a major national objective of US policy.

Nor is this a special case - consider for example stellar populations in globular clusters. A major review by Meylan & Heggie from 1997 has a mere 145 ADS citations. A typical "good" paper in this field might rack up 20-30 citations, because there just aren't that many papers written.
But, van den Bergh, Bolte & Stetson 1990 has 272 ADS citations - a technical paper on relative ages of stellar populations.
Reason why, is that it became relevant to cosmology - it was one of papers that "bubbled out" to point out there was an independent estimate of the age of the universe, or at least a lower bound, and hence a constraint on the Hubble constant and deceleration parameter.
And it was papers like that, which showed stellar population ages were inconsistent with the simplest CDM cosmology (the "stars are older than the cosmos" paradox) which pushed the testing of the CDM models and laid the groundwork for the ready acceptance of Lambda-CDM when additional data indicating that as a favoured scenario became available.

Not by itself, but that is the point. You can't isolated those break-out papers and think only that work needs to be done, because that research in turns rests on a very broad prolonged endeavour advancing our understanding on all fronts, not just those fashionable in this funding cycle.