I answered this
earlierSensible bounds on memory usage per tx or time based rules mitigate or solve this entirely.
If you are putting so much into a tx, then you are exceeding isStandard() rules anyway.
There is currently no solution and you are wrong. XT introduced a per transaction limit (if I'm correct). Core will address this via
BIP 143 ,which is an actual improvement unlike the workaround in XT, and a
limit. However this is node policy, miners can still create a transaction that would take too long to validate and that would harm the network.
I beg to differ, I think you are wrong when you suggest "there is no solution". Even the text of BIP143 describes what can be done - it says its non-trivial, not impossible. Their biggest objection is the hard fork ( as usual). But as I described, this can be mitigated by a refined version of the XT fix.
Bip143 is part of segwit. Its part of a 2000+ line change to implement it. When we are ready for segwit, yeah fine.
XT does not have a per tx limit - its a per block limit. It strives to ascertain rational limits to allow most tx's through but to limit obviously vexatious ones.