When any novel issue arises there’s really no way of predicting what might correct the problem. The only way to find a solution sometimes is by testing many different options to see what might help.
I couldn’t possibly predict if switching schedulers might help. However, if I had a dollar for every time a user questioned a suggestions viability and the suggestion ended up being the solution I’d have a very large stack of money.
If you feel that recent cgroups changes might be responsible, see my post here: