Without inferring anything specific, I'll repeat something I've posted before: descent below MSA in IMC whilst not on a published approach procedure has killed a lot of pilots, some highly experienced.
It would be interesting if you would list some. I think the circumstances are often misunderstood.
The
Graham Hill accident is a fairly good example. It is somewhat misleading to think of it as a homemade approach procedure.
It seems to have been standard practice to receive radar vectors on to final approach and dead reckon from there. Hill's Aztec was vectored to 4 miles and a height of about 1150 ft aal. There was an AOC holder authorised to operate into Elstree with an MDH of 550 ft and a visibility of 1750 m, apparently on that basis. Had Hill stuck to that MDH before positively identifying Elstree then the night would have ended differently.
I would wager that the majority of CFIT accidents on approach involve a loss of visual reference on the visual segment after leaving MDH (for example by misidentifying lights) rather than a collision with a known obstacle on the instrument segment. If you descend below the level of obstacles in your immediate vicinity without visual reference, the results are predictable. It doesn't really matter whether or not the approach is published. The key discipline is setting a minimum and sticking to it.
I guess we'll have to wait until the AAIB report to find out what really happened near Chalgrove that day.