Stop telling the reader that the hundred directly imaged exoplanets were never imaged

Every exoplanet card without a map ended "Not an observation — no image of this world exists.",
and 395b613 wrote that rule into a comment ("Only an exoplanet has never been imaged") and a test.
The NASA Exoplanet Archive flags 102 planets as detected by imaging (ima_flag), all 102 of them in
exoplanets.json: HR 8799 b to e (Marois et al. 2008), bet Pic b, 51 Eri b, AF Lep b, and bet Pic c
and eps Ind A b, found by radial velocity and imaged since. The review read the sentence on the
live pages of HR 8799 b, 51 Eri b and bet Pic b.

The ETL now asks the archive for those names in a query of its own, cached apart from the main
table so the other 6 252 planets stay on the snapshot they were built from, and carries
`imaged: true` on the matching records. exoplanets.json changes by that field on 102 records and
nothing else (compared record by record). Their cards now end "Not an observation — it has been
imaged only as a point of light beside its star, and no map of it exists."; the rest keep "no
image of this world exists". The provenance comment and the texture catalogue's comment say which
is which.

Checked live on :4301: HR 8799 b and eps Ind A b carry the new sentence, Kepler-22 b the old one.

Tests: body-view-model.spec 'says a directly imaged exoplanet was seen as a point of light, not
that no image of it exists'; the old test is renamed 'says an exoplanet the archive does not flag
as imaged has no image'. Unit controls, each failing that test only (1 failed, 840 passed): the
provenance ignoring the flag; the view model dropping it. build.ts now requires at least 95
imaged planets (measured 102); control, the full ETL with the join made on the host's name
instead of the planet's, fails with "Only 0 exoplanets are marked as imaged".

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

tools/etl/fetchExoplanets.ts

@@ -41,6 +41,13 @@ const TAP_URL = `${TAP_BASE_URL}?query=${TAP_QUERY}`;
 // silently wrong rather than visibly broken.
 const CACHE_FILE = `exoplanet-archive-ps-${createHash('sha1').update(TAP_URL).digest('hex').slice(0, 8)}.csv`;
 
+// The planets the archive flags as detected by imaging (`ima_flag`): 102 of them in September 2026,
+// HR 8799's four and 51 Eri b among them, and bet Pic c and eps Ind A b, found by radial velocity
+// and imaged since. Asked for on its own, so adding it did not refetch the table above and move
+// every other planet to a newer snapshot.
+const IMAGED_URL = `${TAP_BASE_URL}?query=select+pl_name+from+ps+where+default_flag=1+and+ima_flag=1+order+by+pl_name&format=csv`;
+const IMAGED_CACHE_FILE = `exoplanet-archive-imaged-${createHash('sha1').update(IMAGED_URL).digest('hex').slice(0, 8)}.csv`;
+
 /**
  * Downloads confirmed exoplanets from the NASA Exoplanet Archive (`Planetary Systems` TAP
  * table), cross-references each host star to the HYG index, and writes `exoplanets.json`.
@@ -52,6 +59,7 @@ export async function fetchExoplanets(stars?: StarRecord[]): Promise<ExoplanetRe
 
   const csv = await fetchTextCached(TAP_URL, CACHE_FILE);
   const rows = parseCsvObjects(csv);
+  const imaged = new Set(parseCsvObjects(await fetchTextCached(IMAGED_URL, IMAGED_CACHE_FILE)).map((row) => row['pl_name']));
 
   let matched = 0;
   const exoplanets: ExoplanetRecord[] = rows.map((row, index) => {
@@ -80,6 +88,7 @@ export async function fetchExoplanets(stars?: StarRecord[]): Promise<ExoplanetRe
       radiusEarth: parseOptionalNumber(row['pl_rade']),
       massEarth: parseOptionalNumber(row['pl_bmasse']),
       discoveryYear: parseOptionalNumber(row['disc_year']),
+      imaged: imaged.has(row['pl_name']) || undefined,
       // The period was already being downloaded and thrown away. With the semi-major axis it
       // determines the host's gravitational parameter, so keeping it is the difference between
       // propagating a planet at its real rate and pretending every host is the Sun.