It is possible to trigger a fuzztest failure in the base_encode_decode by asking it to decode any PSBT that has HD keypaths in it. For example, this one
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
which I took straight from the PSBT test vectors. The reason is that in src/psbt.h we call DeserializeHDKeypaths, which in turn calls CPubKey::IsFullyValid, which in turn asserts that a secp context has been created.
The error appears to be masked on many systems by the definition of instance_of_eccryptoclosure in src/script/bitcoinconsensus.cpp, which defines a static object which contains an ECCVerifyHandle. If you just comment out that line you can reliably trigger the fuzz test failure, e.g. by creating a file crash with the above PSBT, and runnnig
ASAN_OPTIONS=symbolize=0:detect_stack_use_after_return=1:check_initialization_order=1:strict_init_order=1 UBSAN_OPTIONS=suppressions=./test/sanitizer_suppressions/ubsan:print_stacktrace=1:halt_on_error=1:report_error_type=1 FUZZ=base_encode_decode ./src/test/fuzz/fuzz -seed_inputs=crash
