To answer a private-messaged question:

What you are doing here is eliminating any issues imposed by the architectural limits of the queue.   Send units of work across the wire in chunks of convenient and sensible size.   Accumulate them all.   When the receiver hears “over,” it acts upon them as a single atomic unit-of-work.   (And, say, if it hears “forget that,” instead, then it throws-away what it had heard and ignores it, as instructed.)

Don’t let the messages be “unduly large,” such that this conversation becomes a “million-pound elephant” for the queueing infrastructure.   Just send reasonably-sized chunks, as many as may be necessary.