Lightning network and sidechains are not magical things, they're orthorgonal and not alternatives to relay improvements. But since you bring them up: They're both at a _futher_ level of development than IBLT at the moment; though given the orthogonality it's irrelevant except to note how misdirected your argument is...
I appreciate the further detail that you describe regarding the relay situation. So it seems that you are not optimistic that block propagation efficiency can have a major benefit for all full-nodes in the near future.
Yet block propagation overhead is probably the single biggest argument for retaining the 1MB as long as possible. To cap volume growth to a certain extent.
I'm not sure about that. My opinion is that the biggest argument for retaining the cap (or, at least, not introducing a 20x plus exponential increase or something like that) is the burden on full nodes due to (cumulated) bandwidth, processing power and, last but not least, disk space required. Of course, for miners the situation may be different - but from my own point of view (running multiple full nodes for various things but not mining) the block relay issue is not so important. As gmaxwell pointed out, except for reducing latency when a block is found it does only "little" by, at best, halving the total bandwidth required. Compare this to the proposed 20x or 8x increase in the block size.
Bandwidth is always a much bigger concern than blockchain usage on disk. TB disks are very cheap, v0.11 has pruning.