mirror of
https://github.com/lightningnetwork/lnd.git
synced 2026-08-18 13:07:58 +02:00
c2319b17a2
In this commit, we add the control the first hot-load sweep was missing. Warmup payments are real payments, so they hand the router knowledge and drain the network the scored batch then has to use, and those two effects pull in opposite directions: across the mainnet corpus every router got monotonically worse as the warmup grew, until at four hundred warmup payments the whole field collapsed to a twenty-two percent success rate and lnd led on objective purely by abandoning a dead network faster than anyone else. That measures depletion, not the value of a warm cache. A snapshot taken before the warmup and restored after it holds the network fixed while varying only what the router knew when it started, which is the thing a served weight cache actually offers: the knowledge without the spending. Both arms are worth running, since the draining one is the honest model of a node that really did make those payments, but only the restoring one answers the question we asked. |
||
|---|---|---|
| .. | ||
| commands | ||
| lncli | ||
| lnd | ||
| routesim | ||