I've been testing Metaplugin 2.3.8 fairly extensively and it seems like the problem is 'almost solved'. That said, the fact that it appears less frequently arguably makes it more dangerous since you have a much greater chance of becoming complacent about this problem in thinking it's 'mostly solved'. But that said:
I can get Nebulas to recall at 48KHz in this version almost every time. The one time they seem to not recall correctly is when Pro Tools 9 is started for the first time. Something about the timing of starting up the DAW must interrupt the DDMF loads. If PT is already opened and I open a session file, the Nebula instances all come up at 48KHz whether or not you instantiate the instance by opening the GUI.
I tested this by instantiating a Nebula compressor and turned the threshold way down (so the signal is audibly much quieter); saved the project; then when I reopen the project that compressor seems to be doing its thing before, during and after opening the GUI itself.
I've confirmed this behavior on two different systems, one on Lion and one on Snow Leopard, using PT 9 (Metaplugin wrapped into FXpansion of course). This is all using Nebula .492 (I refuse to even so much as download a .60x installer until they're actually fixed, 601 and 602 wreaked enough havoc on my systems to be pretty gun shy).
Because there are still some percentage of times it is not reliably working my feeling is that it still cannot be trusted and every instance must be reloaded for safety. But it seems some changes have occurred in how Metaplugin works that have affected this problem.
Last night my fears were confirmed: using the working methodology of having PT already opened I had a session with 6 Nebulas and each of them failed to instantiate correctly. This is probably 1 out of 50 or so tries that it failed. I wouldn't trust Metaplugin without checking every instance (although most of those checks will probably reveal everything is working fine).
ETA: To confirm that it wasn't something odd with the session itself, I re-opened it several times after that and the instances all loaded up fine.
I haven't read this entire thread but I use the latest version of Nebula Server 64 bit Mac OSX with the latest version of Metaplugin in Pro Tools 11 and everything works fine in 44.1khz. It recalls and opens every single time. Haven't tried 48khz or 88.2khz. And there's a problem in 96khz I recently wrote about where Nebula doesn't distribute the processes evenly throughout all the cores, but I don't think this has anything to do with Metaplugin.
Email came from DDMF today of a new version of Metaplugin:
"the new version fixes the infamous sample rate bug in Nebula and lets you automate parameters from within a single plugin's UI..."
I'm seeing the exact same performance in 2.4 that I saw in 2.38 or whatever I was using most recently: most of the time Nebula instances are loading correctly. The problem seems to be -whenever Pro Tools (this is 9, RTAS wrapping of Metaplugin, btw) is started for the first time, the first session to load will NOT load Nebulas correctly, they all come up as 44.1 (blinking arrow in my case since my sessions are not 44.1). This is whether you start Pro Tools then open a session or initiate Pro Tools opening by opening a session. -since there was at least one random failure not related to Pro Tools loading in 2.38 (and I never figured out what caused it since it was, thus far, a one time deal), I have no reason to believe at this time that such a risk has been mitigated.
Long story short: 2.4 is working exactly the same as 2.38 did, much better but still not good enough to avoid having to train anyone and everyone related to using Nebula about its possible failures. I'll let DDMF know as well.