You have a 24v AC transformer which you say may go up to 28v AC at low current.
When you rectify that, you'll get a DC voltage with a peak of 28v x 1.414 = ~40v , MINUS the voltage drop on the two diodes in the bridge rectifier. Depending on what bridge rectifier you have, that could amount to ~ 1.5v..2v so your peak will be at around 38v.
Remember this is peak voltage, you'll still need some capacitor to smooth things out ...
Anyway, even if you go with a peak voltage of 38v and you'd add let's say a couple diodes to drop to 36v, it would still be a risky idea to use an IC that accepts maximum 36v
You could add a resistor in front of the bridge rectifier as a sort of minimum load, let's say something dissipating half a watt or so, if the total power consumption is not a problem.
If the current is less than 1.2A, a nice and cheap (~0.65$ if you get 100, ~0.85$ in 10pcs) I could recommend would be
AOZ1282 :
http://aosmd.com/res/data_sheets/AOZ1282CI.pdf Absolute max ratings is 40v, the recommended max is 36v... Runs at 450kHz so you can use smaller inductors and at the same time it's not so high that you have to be super strict about layout. You only need inductor and freewheeling diode, mosfet is integrated.
Yeah, in the layout I made in paint, i sort of imagined that you'd have all the right side as ground (a _| shape in a sense) and all the bottom layer could be ground as well (and use VIAs to connect the two) and maybe have a sort of star ground on bottom right corner.
Vias would be OK for the feedback resistors, but I'd like a thicker ground trace going to the chip itself. Some traces (like the one between C16 and C15) could go on the other side through vias breaking that bottom ground fill, if it would allow a thick ground trace to reach the IC.