first you must rule out that only 1 MS is the cause of the problem. Check that the volume of data is about the same than previous days. My idea is that your cell has a low traffic, usually, but on that day 1 guy is heavily using EGPRS. Therefore just one person modifies the whole QoS stats for that cell.
then, check if the other cells in the BSC also face the same problem.
if yes – > probably an Ater congestion
then it starts to be interesting. either you modified how the traffic is allocated on your TRX (pref_mark, frequency hopping, MAX PDCH, etc.), or one TRX is faulty.
In first case – check all parameters changes that occured lately
In second case – restart your TRX’s
Those are the only ideas that comes up, but the solution might not be here. Starting from there, however, I’m sure you’ll find other leads.
For info, MCS is chosen based on TRX capabilities, radio quality and Ater load.