I was thinking you meant a delay of 3 months and maybe I was being hyperbolic but the point stands regardless, anything past 6 blocks is a vanishingly small likelihood if the support is as low as you think. Chainsplit risk is zero barring the proliferation of a URSF client. There are two possibilities:
1. BIP 110 fails immediately (10 noncompliant blocks after activation makes the hash power contest almost unwinnable for BIP110)
2. BIP 110 has moderate to high support making it the main chain.