Avoiding this issue would be a pleasant side effect of encryption. Since encrypted data is indistinguishable from noise, it wouldn’t matter if the underlying format has specific bit patterns that mimic a signal pattern.
1: First, most SONET customers at the time weren't actually buying the whole rate line. (Say you have an OC12 coming into the building, are you actually purchasing an OC12c worth of capacity, or are you buying a DS3/STS1 which is only occupying 1/12th of the line, and the carrier just dropped a larger circuit for their own convenience?) So because of byte interleaving, 11/12 bytes (or whatever) are good, and even if you somehow synchronize with the scrambler, you can't cause enough ones in a row to piss off the laser, or enough zeroes in a row to confuse the receiver.
2: Second, the payload framing means that, even if you are buying a whole OCxC worth of transport, your options are still very limited. I'm rusty on this but I think you only get 87 bits in a row of payload before another framing structure ruins your day. I still feel like it might be possible, but it wasn't within my reach given the test pattern constructor built into the ASA-312.