Core Lightning v0.12.0 through v26.06.7 let a peer open a channel on a funding outpoint that already belongs to another channel, so the two share a v1 channel_id, causing the node to crash-loop on every restart.
Eclair v0.14.0 and earlier have a TOCTOU race that can cause channel actors to be leaked, and repeated execution of the race leaks memory until the JVM crashes or the garbage collector pegs every core.
Eclair v0.13.1 and earlier still accept the zlib encoding type that BOLT 7 retired in 2022 and inflate it with no size limit, so a flood of zlib-encoded channel queries disconnects a node's peers, stalls its API and block processing, and drives the JVM out of memory within minutes.