The various embodiments described throughout this disclosure are advantageous in that a buffer state may be monitored by dynamically sampling the buffer. Hereby, it is made possible to improve the congestion control. A first network node (a.k.a. detection point) creates a congestion notification message in response to determining a condition indicative of congestion. This congestion notification message is transmitted to one or several second network nodes (a.k.a. reaction points). Based on the received congestion notification message, the one or several second network nodes may compensate for a detected congestion by adjusting one or more of its parameters based on received back-off information including e.g. suggested back-off time, suggested back-off rate, and/or suggested ramp-up time. Hence, a second network node may adjust i) the time during which it performs back-off, ii) the rate at which back-off is performed, and/or iii) the ramp-up time for the back-off. Upon adjusting one or more of its parameters, it is possible for the second network node to adaptively adjust its behavior in dependence of a condition indicative of congestion detected by any first network node in the network. This way it is for example possible to adaptively reduce PDCP transmissions/retransmissions in the network at appropriate times. As a result, the transport network may operate more efficiently. As a result, also the user experience may be improved.