I noticed that the the latest BIP 134
<https://github.com/bitcoin/bips/blob/959fecc15bdad070afa63455468b1dba54655fa6/bip-0134.mediawiki>
now supports SIGHASH_SINGLE and friends.  However, this support seems to
reintroduce some quadratic hashing behavior because it calls
<https://github.com/zander/bitcoinclassic/blob/9c688c6d3866890f16a36aaea453e8bdd43c1266/src/script/interpreter.cpp#L1186>SerializePartialTransactionv4
per non-SIGHASH_ALL input
<https://github.com/bitcoinclassic/bitcoinclassic/blob/9c688c6d3866890f16a36aaea453e8bdd43c1266/src/script/interpreter.cpp#L1186>.
In particular, if each input in a transaction has one SIGHASH_SINGLE
CHECKSIG operation then the total amount of hashing done for the
transaction will be quadratic in the number of inputs.  While amount of
hashing is not as severe as with the SIGHASH_ALL case, the amount of
hashing done is still non-linear.