node: fix panic in error cases, fix #2970

internalErr has some buffer to accept multiple errors, but it only processes
one of them, so if we have some cascading failures with multiple components
reporting an error some of them will end up panicking because of closed
channed. It's not very critical since the node is to die anyway, but it's
nothing good either.

This particular case shouldn't really happen after #2996, but leaving this
channel open is better than panic in case some other events lead to multiple
sends to it.

Signed-off-by: Roman Khimov <[email protected]>

CHANGELOG.md

@@ -27,6 +27,7 @@ attribute, which is used for container domain name in NNS contracts (#2954)
 - Overriding the default container and object attributes only with the appropriate flags (#2985)
 - RPC client reconnection failures leading to complete SN failure (#2797)
 - `meta.DB.Open(readOnly)` moves metabase in RO mode (#3000)
+- Panic in event listener related to inability to switch RPC node (#2970)
 
 ### Changed
 - `ObjectService`'s `Put` RPC handler caches up to 10K lists of per-object sorted container nodes (#2901)

cmd/neofs-node/main.go

@@ -173,7 +173,6 @@ func wait(c *cfg) {
 	select {
 	case <-c.ctx.Done(): // graceful shutdown
 	case err := <-c.internalErr: // internal application error
-		close(c.internalErr)
 		c.ctxCancel()
 
 		c.log.Warn("internal application error",