yeah ..I have noticed that current xephyr is an issue...
...I've been trying to "think round corners" (so-to-speak) lately ...
on a few things...that being one...

from my pov...
whatever we can do to get past that "ancient" (ubuntu vm) wall as you so aptly put it...
to bring maemo development into the present ..(if at all possible)..in whatever ways we can.
I don't think any ideas or methods should be off the table.
since anything that works, partially works, or doesn't..
may work.. if approached in a different way..applied in a different fashion
your new jessie toolchain is a fabulous improvement .
New approaches, new ideas are welcome
Worst case we just end up with some live distributable rolling-release iso's ..
containing vmbox'd ...
antiquated 'buntu os...
containing scratch...
containing sdk's...
containing everything we need...
which is truly a great amount of bloody luggage to carry around...
like a freakin' Ukrainian doll-within-a-doll-within-a-doll...etc.
It's getting ridiculous.
(qtsdk 1.1.1 is a superb, elegant and sublime piece of work in compare...but that was done by team(s), a company, money, and a decade of work leading up to it...behind them)
So yeah...the more I look at the picture I'm dealing with...(and actually we all are dealing with) ...particularly concerning making thing easier, simpler, more current and updatable...the more I'm looking at obtuse and oblique methods of approach to reach that goal.

In short you could say..
I'm " wingin' " it in a squirrel-suit ..
that the disgruntled -ex I dumped years ago...
sewed for me last night ...
off too-short a building ..
I'm not sure how anything is going to turn out. or if it's going to turn into anything I was aiming at in the first place.
One thing that is a certainty...
I will land on the ground...
well "land" may be a generous word.