Replication via binlogrouter temporarily blocks the REST-API

When the slave thread starts maxscale hits high usage on a single node leading to very slow maxctrl returns. For this customer it leads to flapping on health checks.

The slave thread connect appears to completely take a core on occasion or just spike high, why is this hitting a bottleneck for 1 connection?

##On replica

MariaDB [(none)]> stop slave 'ext_repl';start slave 'ext_repl';

##Usage observations 

 670282 maxscale  20   0 4586296  51076  14948 S  56.1   0.1 558:38.14 maxscale

##Seen delay in maxctrl return

Wed 26 Jul 2023 08:59:34 PM UTC

Wed 26 Jul 2023 08:59:36 PM UTC