Prophylatic reconstruct before xfer?

[wcooley at nakedape.cc]

I have a bunch of older 2.2 backends that I am migrating mailboxes off of,
on to new 2.3 backends. I am starting with my dead accounts (which should
really be deleted, but for various reasons aren't), which means these are
among my oldest and cruftiest accounts.

For reasons that I am only sometimes able to ascertain, some xfers fail or
they partially fail (a subset of folders will transfer, but not the INBOX
and possibly some other folders). Sometimes the xfer will just hang,
forcing me to kill it when I find it hours later. (On my to-do list is to
put in a reasonable time-out, so it doesn't waste the whole night.)

I am looking for ways that I can prevent these failures or hangs, so I am
considering a few things, among which is unconditionally running a
recursive in-band reconstruct before attempting the xfer. Would this
increase the risk of something getting corrupted with these old 2.2.12/13
backends or should it be safe?

I have noticed that in many cases the reconstructs return an "Operating
System Error" on otherwise fine mailboxes (including a test mailbox that
doesn't have more than a handful of messages and was recently created).
Why would that be? I am considering ignoring the return value of the
reconstruct command, which makes me even more uneasy about it.

Wil