The campaigns grid renders up to 200 CampaignCards at once with no
virtualization, and each card eagerly ran useCampaignDonations: an
Esplora /address lookup per card (polling every 30s via refetchInterval)
plus a /tx verification call per kind-8333 donation receipt. On first
paint that's hundreds of /address calls and potentially thousands of
/tx calls in one burst, then 200 more /address calls every 30s forever.
Once mempool.space starts returning 429, the failover client retries
each call across every configured endpoint, cascading the rate-limit to
all of them.
Gate the donation lookup behind an IntersectionObserver so only on-screen
cards talk to Esplora, and drop the per-card 30s polling. The detail page
(a single instance, not a grid) opts back into the 30s live refresh via a
new refetchInterval option.
- Add useInView hook (once-only, rootMargin pre-arm, IO-less fallback).
- useCampaignDonations gains enabled + refetchInterval options; default
no polling, longer staleTime; treat 'not enabled yet' as loading so
off-screen cards show a skeleton instead of flashing '0 raised'.
- Gate CampaignCard and CampaignInlinePreview on visibility.
- CampaignDetailPage opts into refetchInterval: 30_000.
Alex Gleason
a8b2fe5ddf