Hey wossi,
I think what you're looking for is doable. I'll go over the problem and then two possible solutions.
The glow effect is based on a Gaussian blur, and this is where the problem arises.
Normally using two Gaussian blurs at right angles will produce a radially uniform (smooth corners) Gaussian blur, (a very cool and special mathematical property of Gaussians). Unfortunately this special property of perpendicular Gaussian blurs only works when they're combined multiplicatively. (Using a horizontal blur to blur the result of a vertical blur is effectively multiplicative.) Unfortunately, adding or "screening" them together, which is what Construct 2's "Glow" effects do when stacked in series, does not preserve this special Gaussian behavior. So the result of perpendicular "Glows" is not radially uniform.
So, stacking a regular H-blur and V-blur will give you the smooth corners you want, but it won't glow.
Stacking H-glow and V-glow will glow, but it will give you sharp corners.
What you need to do is create a blur-glow system within Construct2.
With a canvas object (maybe):
Disclaimer: I've never used the canvas plugin (3rd party I think), but if you can paste objects or layers into it then it might make a glow effect semi-simple.
1. Place a layer below the foreground objects and above the background.
(May need to set the scale and scroll rate for this new layer to "0".)
2. Stretch a canvas object across that layer to fill the window.
3. Every tick, clear the canvas to transparent, and then paste into it all the objects you want to glow.
4. To the canvas, add the effects "horizontal blur" and "vertical blur", then add the "Screen" OR "Dodge" effect.
Done.
Using only native C2 stuff - (no canvas):
Without the canvas it is still doable, but it takes a little more work and clutter.
For a given (source) object that you want to glow, you create a sprite with an identical image, and set it to match the position, angle, and animation frame (if animated) of the source object. This copy object must have a Z-index below the source, which you can set with events. To this copy object, add the effects "horizontal blur" and "vertical blur", then add the "Screen" OR "Dodge" effect.
"Dodge" vs "Screen":
Dodge and Screen are both glow-ish style blend modes, in that they will make an object transparent, such that it brightens the backdrop.
"Dodge" (sometimes "linear dodge") is literally just "Addition". It's that simple.
Blend result = BG + FG.
"Screen" is exactly the same as "multiply", but upside-down. Where "multiply" pushes all brightness values towards black, but never past black; "screen" pushes all brightness values towards white, but never past white;
Blend result = invert ( invert BG * invert FG )
i.e.
Blend result = 1 - (1 - BG * 1 - FG ), where black is 0 and white is 1.
Practically speaking why use one vs the other?
"Dodge"/"Addition" is good when you want a vibrant intense look, with the potential for blown-out overexposed areas. The results for additive blending can be brighter than white, so they will be clipped to white. This is good for fire, electricity, and other intense eye-burningly bright effects.
"Screen" is good when you want soft glow, or bright-but-smooth looking glow effects. Screen can produce pure white in cases where the FG or BG is pure white, but it will never blow-out or overexpose areas, because the mathematical result of "screen" blending is always between 0 and 1, so no clipping can occur.
Hope that helps.