Safety layer · measured in public
Speed & uptime
How long a coin check takes and how often Toll answers at all, from the log every check writes. p50 is the middle ask, p95 the slowest one in twenty. A cold check is a coin nobody had asked about, read fresh from the chain.
Nothing to show
How it is measured
One ask is one check request, with every “still reading” poll of it folded in: its time runs from the start of its first request to the end of its last. Its outcome is the last answer: whole (the full check), still reading (counted as answered, not as whole), busy (our chain-read budget for the minute was used up) or failed.
Answered is every ask that was not busy or failed. It is the share of asks Toll answered, not a ping of the server. A coin not read yet gets a first look within about 3 seconds (marked INCOMPLETE, never a clear) while the full check reads; that first look counts as still reading here.
Times are in UTC. Our own internal calls are left out. No coin, wallet or visitor is part of these numbers.