I don't personally believe it's covid causing delayed diagnosis of other underlying diseases because it's a persisting effect and appears broadly linear to the 5 year trend including that bizarre knee in the data. (Which may be the weight of this trend in the 5 year model I guess. They says it's the effect of the ONS office closure in May/June, so perhaps it's just data handling effects overall)
I'd be much more concerned it's a post covid mortality statistic which may still be under-reported worldwide.
Could be one of those "wait and see" things. Cause of death and prior covid, long covid and vaccination status will be important to understand it but a problem here will be the sheer ubiquity of covid, and vaccination.
A single event like extreme heat wouldn't evenly distribute deaths in an ongoing trend surely?