Roman's 100x field is a pipeline bill, and August 30 is when it comes due

Matching Hubble's resolution across a field about 100 times larger is, arithmetically, about 100 times as many pixels per pointing, assuming comparable sampling per resolution element ^[2][4]. That ratio is not exotic for a survey camera, but it does not land on the optics. It lands on whatever downstream of the focal plane has to calibrate, store, difference and catalogue the output for five years ^[3].

The science makes the pixels harder to discard. Tracing structure by measuring subtle gravitational distortions of light ^[4] means per-object shape fidelity across the whole survey, which is exactly the kind of measurement that aggressive on-board compression damages. The planet search compounds it in the other direction: scouring hundreds of millions of stars ^[8] for transits and for microlensing alignments ^[9] means the same fields get imaged repeatedly, so volume grows with cadence and not just with area.

Then the count. As many as 200,000 new planets ^[5], against roughly 6,300 confirmed today ^[6], is a catalogue about 32 times larger ^[1], and across a five-year primary mission it works out to an upper bound near 40,000 a year, or about 110 a day ^[2]. Nobody vets 110 objects a day by hand. Whatever process turns a candidate into a confirmed planet becomes the rate limit long before the telescope does.

The microlensing haul is the part worth protecting. Roughly 1,000 detections is about half a percent of the total ^[7][3], and it buys the objects the transit method cannot reach: planets less massive than Mercury, and free-floating worlds with no star at all ^[11]. As Louisiana State University's Matthew Penny puts it, the method does not require the planet to complete an orbit before you see it ^[10]. That is a signal you catch while it is happening, which makes visit scheduling and processing latency scientific parameters rather than operational ones.

Roman also inherits Webb's address, an orbit around L2 nearly a million miles out, and works mostly in the infrared ^[12]. Carnegie Mellon astrophysicist Rachel Mandelbaum draws the contrast plainly: Webb is optimised for the very distant universe over a tiny patch of sky, while Roman takes in a large area at once ^[13]. Deep and narrow produces datasets you can argue about object by object. Wide and repeated produces datasets you argue about statistically, and the statistics are only as good as the systematics the pipeline removes.

"As soon as August 30" from Kennedy Space Center ^[1] is a floor, not a date. The asymmetry matters for everyone downstream: launch dates move late, so ground software, calibration plans and follow-up time all had to be ready for the earliest possible date and then carry the cost of waiting. Project scientist Julie McEnery calls Roman intrinsically a discovery machine, expecting rare and surprising things ^[14], and says the most exciting result may be something nobody can imagine now ^[17]. That is a candid description of a selection problem. The cuts that decide what survives the front of the pipeline are being written now, by people who cannot yet describe the anomaly they most want to keep.