From what I can see of DLM in the past 1-2 years on fast path recovery is that on a first incident it will now recover very quickly within 1-2 days, if you waiting 20 days then its probably not a first incident and I agree with openreach that a repeat incident should take longer to recover. More speed is nice, but not when it results in instability.
The issue as I see it is the lack of manual override mechanisms in place rather than the speed of recovery, and of course how quick DLM reacts (often reacts after the event).
In the case of manual override where the customer gives their consent to risk the line been unstable, its fine to go ahead and force the issue, but with automated recoveries the customer has not given their consent to compromise stability which may be of real importance to them.