One error flood, one issue.
Epure stores a sample, counts every repeat, and the next bug still shows.
Once you’re in: 14 days on Pro limits, no card. Your Sentry SDK stays.

Bottom line
Epure stores 100 events a minute per identical error and 150 per throw site, then counts the rest. A loop stays one issue. A new error still opens its own.
| Dimension | Payloads stored | Occurrences counted | A new bug during the loop |
|---|---|---|---|
| Store every event | All, until the quota is gone | All | Lost once the quota is spent |
| SDK sampleRate 0.1 | 10%, rare errors included | Only the 10% sent | Sent 1 time in 10 |
| Epure spike valve | 100/min per error, 150/min per throw site | All | Its own issue, its own budget |
Rates from crates/server/src/pipeline/spike.rs in Epure v0.2.1. Same on Pro, Scale and self-hosted.
Without a valve, one TypeError spends the quota. The next bug sits underneath.
Updated · Epure v0.2.1 · Beta
One loop hides the next bug
Storing every event spends the quota on the loop. Sampling in the SDK thins the new bug too.
Without a valve
One loop spends the quota
One TypeError spends the quota. The new bug sits underneath.
With Epure
One issue, exact count
The loop stays one counted row. The new bug opens its own issue.
How the spike valve works
Identical errors share one budget. A new error gets its own.
01
Fingerprint the repeat. Identical errors join one issue and share one budget.
02
Store a sample. Up to 100 events a minute per error, 150 per throw site.
03
Count the rest. The count stays exact. A new error gets its own budget.
The limits
Same rates on every plan, from pipeline/spike.rs. Duplicate SDK retries drop for 7 days.
pipeline/spike.rs
100/min
Stored per identical error
pipeline/spike.rs
150/min
Stored per throw site
durable ingestion
7 days
Duplicate SDK retries dropped
v0.2.1
Every plan
Same rates on Pro, Scale and self-hosted
Check it yourself
On staging: one error 500 times, then a different one. Expect two issues. Within a minute, Epure stores 100 payloads and counts the rest.
// Staging only: one looping error, then one new error.
for (let i = 0; i < 500; i++) {
Sentry.captureException(new Error("loop: cart total is NaN"));
}
Sentry.captureException(new Error("new: checkout button missing"));The count shows what reached Epure. Sentry's Dedupe integration drops an event identical to the one before it, so a tight loop of the same exception can arrive as far fewer events.
What it doesn't do
- Fix the loop
- Your SDK still sends every event. The fix lives in your code.
- Keep every payload
- Past the budget, Epure keeps the count and drops the payload.
- Vary by plan
- Pro, Scale and self-hosted use the same rates.
- Track traces, replays or logs
- Epure drops them at ingest.
Spike valve questions
What happens to events past the spike valve limit?
Does an error flood hide new errors?
Does the spike valve change my bill?
Is it the same as Sentry's spike protection?
Do I get one alert per event during a flood?
Keep the next bug visible.
Loop one error 500 times. Throw a new one. Expect two issues.
Once you’re in: 14 days on Pro limits, no card. Your Sentry SDK stays.