You should worry about the opposite: An overhand knot on nylon webbing will get really tight and hard to undo once no longer in use. That's why I rather use a figure-eight in this scenario (double bowline would be even better for undoing, but a bit more cumbersome to tie). This setup is somewhat standard practice for abseiling here in Switzerland, where you tie a 120cm sling to your harness, make a knot at about 2/3 length and attach your backup (prusik) to your harness, the braking device (e.g. tuber, eight) to the lower loop of the sling (same position as in your image) and the biner for self-securing in the outmost loop of the sling (undone when abseiling, attached when at belay-station). So plenty of anecdotal evidence that they definitely don't slip.
And if you don't trust that: Would a failure be catastrophic? No! The loop would just extend to it's full length, you are still clipped into the "main" loop so you wouldn't get disattached. That's also why this setup is ok even with dyneema slings. Again anecdotal evidence says by not shockloading, they just pull tight (and in case of the narrow slings very annoyingly, potentially almost irreversibly tight). Tests have shown that under high stress, they do creep.
It won't come undone, it will get really tight. And even if it did (it does not), that wouldn't be a catastrophic failure.
As to breaking of the nylon webbing (not in the question, but just to be thorough):
The rule of thumb is to assume a knot weakens the nylon by 50% (common climbing knots lie below that). A nylon sling as a loop typically has 22kN breaking strength, i.e. 11kN on each strand (as a ring is always equally loaded, as long as you don't fix your biner with a clove hitch or something). Now one strand has a knot in it, meaning now the breaking strength of the knotted strand is 5.5kN and the weakest strand breaks first, so 11kN for the entire sling.