Just how stable do printers need to be?

Having printed with the Prusa Mendel, the Mendel90 and a printer of my own design (also a Mendel-style one), I came to wonder at which point an unstable frame limits printing accuracy.

I think it is well known that at certain speeds, the Prusa Mendel's X axis will shake up due to lacking rigidity of the construction. The Mendel90 solves this issue completely due to a superior frame construction.
My printer was not so well thought out and will, like the Prusa Mendel shake in X direction. I am already planning the next iteration, but got stumped by the fact that M8 threaded rods are in fact quite flexible.
Here is a photo to show an issue I am having:


This is a prototype for the gantry of my second frame. The long section to the right is not supposed to be there in the final build, but I used parts intended for a heated chamber to quickly slap together a proof of concept.
The left side with the handle is quite stable and won't wobble unless I am pushing hard against the sides. This will be the top of the gantry. The lower (right) part has to make do with a single threaded rod because it acts as a pivot - the gantry is supposed to fold down for transport. The Y axis will be attached to this part and clamped tight with wing nuts.
Now here is the problem: Even though the rest of the gantry is stiff and stable, the single M8 rod is flexible enough that, when clamped down close to the gantry sides, it will still bend under light loads applied to the top of the gantry due to the leverage. We are speaking of about 1cm of M8 rod that is between the gantry sides and clamps for the Y axis that will flex.
Currently, I plan to use the base of the gantry as feet for the printer. This way, side loads might get distributed to the surface below the printer before it can shake up. I have tried this out and it indeed improves stability a lot, but it still sways slightly when I press against the frame.
So, can anyone with more experience can give their 2 cents about how unstable a printer can be before print quality gets affected (at infill speeds of 60-90mm/sec)? My current printer's output is quite decent against all logic because I can easily see the hot end shaking after decelerating from a travel of 10 mm (at 90mm/sec, 2000mm/sec^2 acceleration) already and the construction is crap, so this adds to the confusion.

My printer. Obviously, I was quite silly to use sheet metal like this.


Also, an XY table, Z head design (like on the Cupcake for example) seems nice for mounting heavier extruders and other experimental tool heads, but obviously, the printed part will move a lot more. Is this trade-off worth the benefit of being able to experiment more with heavy tool heads? Or is there some enormous drawback aside from greater space requirements I am not seeing (as only few printers use this design)?

you also need to look at material thermal expantion properties as well. for example your metal frame may be more rigid, but it has thermal expantion during heat up. wood such as in the cupcake cnc swell in moist weather,and shrink in winter with low humidty, causing at least seasonal adjustments to the layer hieght. so how accurate are we talking about in the end? materials used, temp and humitidy consistency, and also rigid construction.


tentioning seems to be the way to manage these variations, with homing when device is in its working conditon, heated up.

Argh, you are right! Up until now I sort of ignored thermal effects on the frame.
Somewhere I have read that an aluminium heat bed will expand 1mm sideways when heated to 100°C, so the aluminium extrusion frame I was considering for the second build might have some troubles in a heated chamber...

I was more concerned about mechanical stability and the effects of a slightly wobbly frame on the print, though.