[sr-dev] t_suspend() in TMCB_ON_FAILURE?

Hi guys, since tm is a bit, er, complex and breakage is not immediately obvious, I'd rather ask: Is it safe to call t_suspend() from the TMCB_ON_FAILURE callback? The idea is to always delay the transaction until there either is a 200 OK or it times out. In other words, does this work: static void failure_handler(struct cell *t, int type, struct tmcb_params *params) { if (params->repl != FAKED_REPLY) { tmb.t_suspend(params->req, &t->tindex, &t->tlabel); } } If it doesn't work, any better idea? Regards, Martin