This only makes sense when looking at the copper pour, not at the schematic. The point of this is to keep the power ground isolated from the regular ground, so that the pour areas are only connected at a single point carefully chosen layout-wise, for EMC purposes. The point is to keep the high currents of the SEPIC hot loop away from your general ground.
Not directly related to your question, here's some notes from a "concerned citizen":
However, following TI guide lines only gets you so far. I had multiple competent power designers working with LM3478, many years back by now (disclaimer: I'm not a power electronics engineer myself but was the lead engineer/project manager). Everything worked fine in terms of efficiency & stability etc working in a wide range between 11-30V for a stable 24V out. The only problem was radiated emissions, which we never could manage to keep down no matter what we tried. On top of that, TI tech support was and remains non-existent - truly horrible.
As with many parts coming from the old "National" side of TI, the solution was to get rid of it and switch to another brand. We went with LT3759 instead, again following guide lines from the manufacturer, same EE designers did the new design. And then all EMC problems were instantly gone. At which point I'd quite confidently say that the problem was the IC and not the design. LT also had great support back then (nowadays AD).
PGND
? \$\endgroup\$HVGND
. What is thePGND
net connected to? \$\endgroup\$