LMR33630: Short out in buck-boost inverting configuration

Part Number: LMR33630
I'm trying to use the LMR33630ADDA in a buck boost inverting configuration, as suggested by the webench configurator. Used almost the same components suggested (only slightly less output caps since output ripple is not an issue in this application. The requirements are 8-18V to -12A with about 1A average output current. The implemented circuit is https://ibb.co/6RMvCkF (the enable circuit is not connected, EN is strapped with a zero ohm to VIN; when the enable is to the chip ground of course it doesn't work but doesn't explode either). Problem is: it doesn't work. It works in a sort-of hiccup way down to about -5V then it burns out. This with *no load* (the load is discontinuos and it's applied later). My theory: I've read that this architecture suffers from the dreaded RHP so maybe is not stable and self destructs with no load (also this LMR from the datasheet has a special low load mode, maybe is that)? I've actually never got how the initial biasing occurs since the converter sort-of pulls down its ground itself. Through capacitive charging? some hidden internal pathways? As for the layout I have C43, C44 and C40 right next to the converter to avoid brownout issues. It's difficult to make measures since it burns out after few seconds. Read SLVA257, SLVA059, SNVA856, SLVA317 and similar notes from other manufacturers but no enlightment yet. Any suggestion?