DROP ALERT NOTES
When the slate starts: planning for the alert burst
The first minutes of a big slate produce a burst of drops as markets reprice at once. How to keep your relay readable when everyone moves together.
A Saturday slate starts and the drop stream turns from a trickle into a burst: dozens of matches reprice in the same minutes, and every one of them can cross your threshold at once. The burst is not an anomaly; it is the stream's busiest hour, every week, on a schedule you can read. Planning for it is the difference between a useful channel and a flooded one.
The burst has a schedule
Alerts cluster where matches cluster: the top of the hour when a league's round kicks off, the half-hour line when a slate's first games begin. The live vs prematch page covers how different the two rhythms' volumes are, and the burst is the moment they collide: the prematch board's last moves land as the live markets open.
You can read your own week's burst shape from your log: alerts per minute across the day, and the two or three spikes that repeat every weekend.
What breaks first in a burst
The bottleneck is rarely the stream; it is your delivery. Fifty alerts in a minute are fifty webhook posts, and the platforms answer bursts with their own 429s, which the relay guide covers with its retry rules. A relay that posts one message per alert will spend the burst in retry loops and deliver everything late, which is worse than delivering less.
The reader experience breaks next: a channel that fires forty times in a minute gets muted, and muted channels are where alerts go to die.
Absorb the burst before it reaches the channel
Three settings from the relay guide do the absorbing: one message per event instead of one per outcome, a cooldown per selection so a price that keeps falling does not post every few seconds, and a queue that drops the oldest entry when it fills, because an alert about a price from three minutes ago is no longer news. In a burst these are not optimizations; they are survival.
A fourth layer helps on the biggest slates: raise min_drop for the burst window itself, so only the moves that matter at volume get through, then lower it back after the rush.
Watch the burst, then size for it
Log the burst for a few weekends before you tune anything: alerts per minute at the spike, how long it lasts, how many survived your rules. The number you size for is the busiest minute you intend to follow, not the quiet average, and the burst's own shape tells you which of the three absorbers to strengthen first.
The burst arrives on a schedule, every week. The relay that survives it is the one that expected it.