It was not said by Peter Todd, but Pieter Wuille in his original Segwit presentation: http://diyhpl.us/wiki/transcripts/scalingbitcoin/hong-kong/segregated-witness-and-its-impact-on-scalability/  he was talking about the possibility of a new kind of node between SPV and full node, and it has nothing to do with the full node security during a hard fork. Cypherdoc misrepresented it as " and since one of the stated benefits of SWSF was that it shifts the security assumptions away from miner collusion and non-Sybiling to that of "non censored partially validating SPV node connections" via as yet uninvented/coded fraud proofs in an attempt to scale nodes, this just compounds the problem given that we now know that the p2p network in a SWSF can be parititioned. we can't tell if SWSF is safe."

Your full node would not be downgraded to a "fraud-proof" node in the scenario of a SWSF, AFAIK the code for fraud proof has not even been written, it's until now a mere theoretical construction.