> I've been bitten by this same issue where kill-buffer /
> replace-buffer-in-windows doesn't take available tab-bar tabs into
> consideration as viable replacement windows to restore when quitting the
> last window on a frame.  I looked at 29+, 30, 31 window.el/.c and
> tab-bar.el and I can't find if this was resolved or if there was a recipe
> to avoid this.
>
> I've resorted to wrapping kill-buffer using a custom function, not advice,
> to inhibit deleting the frame.
>
> Do any of you recall if this bug was addressed and how?  If not, may I
> assist in some way?

Sorry, I'm still testing the fix attached below.
Does it work for you?