I bet you do not need an exact answer. PCB capacitance will normally be small compared to device capacitance, and you are smart enough to layout any C-critical nodes as small as possible. You just need an order of magnitude: 0.5pFd or 5pFd?
I was going to say that most insulators run 3 to 10 times the capacitance of a vacuum, but JH gives FRP = 4.0-4.5 which seems fine to me. Thanks, JDB, for the links; and Welcome.
> It is not clear to me if the plate capacitor formula is a good approximation, as one "plate" is much smaller than the other. In the case of side-to-side tracks the area A seems to be rather ill defined.
I bet you can solve this near-enough with over-simplifications.
When one plate is so wide and the other is infinite, clearly the area under the trace adds a lot, the area far from the trace is negligible. Say the trace acts twice as wide as its actual width.
For side by side traces: you have the edge-area which is small but also small-distance, and the "top" areas which are large but further apart. Compute the edge area and distance, assume the e is about 4, find that capacitance.
For the "top" and "bottom" capacitances: draw a bunch of circles representing the electric field. Find the average path length. Use that and the area to estimate the "top" capacitance, which is effectively in vacuum. Multiply by 4 to find the bottom capacitance, since most of the circles will be inside the FRP. This answer is surely wrong. You can divide the traces into infinitesimally narrow strips and compute each one with calculus. There is probably a simple universal answer for infinitessimal space between strips. But the rough answer is probably good enough for your purposes.
It can matter. At MHz, surface mount has advantages in part because of smaller stray PCB capacitance. But I think if you have to ask, a rough answer will focus your attention on where it matters and where it can be ignored.
You could also build an RC oscillator with CMOS, big R (like 100K), top-accuracy 10pFd caps paralleled with a few inches of trace capacitance. Measure the oscillation frequency, use a knife to cut-out the trace-caps, measure again. The change in frequency is proportional to the change of C. The C is not known exactly: 10pFd parts may be +/-1pFd and you have unknown stray pin and pad capacitance. (In theory you can use the R to compute what the frequency "should" be, and know the true capacitances; CMOS is not that precise and a precision oscillator for very small C is not easy.) Air-wire as much as you can, to stay off the hi-e FRP.