Developers building infrastructure for AI agents often assume wholesale prices drift smoothly. A new analysis from the team behind the eSIM operator kit proves otherwise. By comparing nightly catalogue backups from October 19 to November 18, 2026, they discovered that prices remain static for long stretches before jumping in sudden, synchronized batches. The investigation began when an engineer from APIbase questioned the validity of a ten-minute quote window used in the kit. The team admitted the duration was arbitrary. To find the real answer, they analyzed 18 nightly archives of the eSIM Access catalogue, containing between 1,725 and 1,823 plans. The data showed 402 price changes across 377 plans in 30 days, but these were not distributed evenly.
The Batch Pattern
For the first 17 days of the observation period, no prices changed. Then, the behavior shifted dramatically. In the following 13 daily intervals, seven days brought batches of 16 to 111 price changes, while six days remained completely static. The majority of these changes were price cuts, with a median drop of 14.3%. However, all 46 price increases occurred in a single update window between November 15 and 16, with a median rise of 25.2% and a maximum spike of 88.2%. To pinpoint the timing, the team deployed a recorder on November 18 that checked the catalogue every five minutes. They observed a batch of 17 plan cuts landing between the 16:14 and 16:30 UTC reads. This confirmed that the risk for a developer is not slow drift, but a wholesale price update landing inside a short quote validity window. While the probability of a single quote overlapping a batch is low, roughly 0.4%, it affects all plans in that batch simultaneously.
Handling Stale Quotes
The kitβs logic in esim-rails 0.5.1 handles these jumps by honoring the quoted price only if it remains above the seller's wholesale cost. If a price rise pushes the quote below cost, the system returns a 409 error with a fresh offer. Testing with Node 22 confirmed that the example buyer script correctly switches to another seller or accepts the new price within set limits, preventing losses from stale quotes.
Key Takeaways
- Wholesale eSIM prices change in sudden batches, not smooth drifts, with 30-day data showing 17 days of zero changes followed by rapid updates.
- A ten-minute quote window is statistically safe for most individual quotes but vulnerable to batch updates affecting multiple plans at once.
- The
esim-railskit mitigates risk by rejecting quotes that fall below wholesale cost, returning a 409 error with current offers instead of processing a loss. - Current data covers only one supplier, eSIM Access, and represents catalogue costs rather than actual transaction prices, requiring further observation for broader applicability.
The Bottom Line
Stop guessing your quote validity windows. If you are building agent infrastructure, you must account for batch updates, not just incremental drift, or you will get burned when the wholesale catalogue suddenly rewrites itself.