It could be argued that this is a poor reason to switch from Erlang to C -- this is a problem with how binaries are handled in user code, not something intrinsic to Erlang VM.
Generally, the write amplification will be < 7 (max # of levels), but in practice is probably ~3.5 since it's possible for files/keys to get promoted across multiple levels under some circumstances.
I haven't measured specifically what you asked for, but have watched the compaction system quite closely in heavy production loads. :)
You've clearly never had cancer. I too considered myself a "reasonably healthy person" and was surprised to get diagnosed at 32 with stage 4 lymphoma. I've had it and done the chemo and come out the other side (for now). My cancer is one of the more treatable ones (>80% survival over 5 years), the chemo one of the lesser ones (R-CHOP). Even so it was one of the most difficult experiences in my life.
Trading a probability of longer life for a higher probability of cancer is just stupid. Most cancer treatments have the potential to (dramatically) shorten your life -- i.e. the combined probabilities probably don't work out in your favor.