My guess is that when the failed drive is replaced, the compaction is immediately halted and the remaining un-compacted data on the failed drive is replicated to the replacement. That is the logical thing to do because if the compaction continued anyway, the entire drive would have to laid out again (but in protected background mode) after the compaction. That makes no sense (from my end user point of view).
Assuming that compaction stops with the replacement, a compaction is probably more stressful to the remaining and very critical drives because they are not just being read in order to reconstruct parity and data zones. They are being totally re-written. And a compaction probably takes more time, assuming reads are faster than writes.
And even if the read and write rates are identical, a compaction requires all data on all remaining drives to be read and written, instead of just read in order to compute the zones on the replacement drive.
I don’t have any opinion either regarding what happens if the replacement drive fails. We would hope that that is recoverable, and that is probably the most likely disk to fail, assuming there is more early failure than more or less random failure on seasoned drives.
The reason for my P.P.S. is that, regardless of how it goes about it, and regardless of when you replace the drive, a full relay will result and during that time, depending on what we just discussed, any drive failure could kill the array (assuming single disk redundancy).
However, if you leave the compaction running, intentionally waiting for it to complete before replacing the drive, then any further disk failure is guaranteed to fail the array, and it is guaranteed to take the maximum possible time to complete (longest windows of vulnerability).
If you replace the drive ASAP then it is possible that at least a failure of the replacement drive is recoverable, and there may be less stress on the surviving drives (if the replaced drive is indeed just rebuilt rather than the entire array rewritten in some complicated way).
I don’t see any possible advantage to leaving it to compact, except the slim chance that you might knock another drive loose AND the Drobo does not allow you to recover from that. I think that likelihood is less than all the other possible benefits of quick replacement that would likely reduce all the other risks.
The worst case scenario for ASAP replacement is that despite the replacement, the compaction goes on just as if the drive were not replaced, resulting in no possible benefit to ASAP replacement. Leaving a slight additional risk added by the physical hot swap. But I think that is a very unlikely outcome.
Not well defined probabilities but I find it difficult to concoct likely statistical outcomes in favor of letting it compact. Greater minds might well come up with one 