The latest network write/read is already tracked for the connection_timeout feature. It covers both read and write timeouts and controls how long a client can be idle before the connection is closed.

connection_timeout and net_write_timeout are different things,

The connection_timeout parameter is used to disconnect sessions to MariaDB
MaxScale that have been idle for too long. The session timeouts are disabled by
default.

The net_write_timeout is number of seconds to wait for a block to be written to a connection before aborting the write.

Oh yes, I think we could add a similar mechanism into MaxScale. We already know if network writes are buffered so the only thing to add is to see if they have been buffered for a long time.

We could handle the read timeouts by comparing the last read and write times for a client socket. If a client sends a request the read timestamp will be updated. When data is written to the client socket, the write timestamp is updated. By evaluating write_timestamp - read_timestamp > net_read_timeout, we will know if a client request is taking a long time.