Only a large bulk filter cap is shown. That can fail to
catch the peak voltage delivered by the boost inductor
(that C*V suggests an electrolytic or tantalum cap,
these tend to have dismal ESL, often poor ESR as well,
and self resonant frequencies in the kHz range).
High current implies high inductor current and high
crest voltage at the filter.
Have you put a 'scope to the supply output at all?
Looked at the ripple there under failing conditions?
For that matter I see zero discussion of any attempt
at debugging, only "broke, boo hoo".
Now, with 3.7V in and a "failed" XL6009, how do you
get more than 3.7V anyway? Only can get there if
the pump is pumping. Otherwise it's just 3.7V, inductor,
Schottky, load (so maybe 3.3V-ish). Just what does
"failed" mean, and where is this purported high voltage
(=?) coming from?