johnh113 - yeah, there's not much we can do about the slow startups with a list that big; the way the system is currently designed it has to check every flashcard in your database before it starts a session, so with 25000 cards it inevitably takes a while. In a future version it should be possible to optimize for cases when you've got a lot of cards but only a few in the categories you've selected, which would both reduce the memory requirement and speed things up considerably, but that would introduce way too many new bugs to consider it for 2.0, so for now the best way to improve performance is to reduce the number of cards you have.
It shouldn't be behaving this badly, though - we actually used your flashcard file for stress-testing quite a bit here (it was the largest one anyone sent us) and were able to successfully start sessions using all of your cards quite a few times. (on a 680, and they didn't take 40 seconds either) Where did you tell Pleco to store your flashcard databases, in internal memory or on an SD card? Are you sure you successfully deleted the old file before your second import attempt? One thing you could try would be to go to the Search Paths section of Preferences, set it to "Create new user DBs in" PDB instead of internal/external, then delete the old database and try again - it's possible that the PDB storage system might work better than the internal flash one in your particular case. (it's what we were using for most of those tests, anyway - the internal flash one was added later in response to widespread issues among Palm TX users with the PDB system)
estudiando - I think these problems actually have more to do with the greater complexity of PD2 searches (more dictionaries to search, plus more checking / sorting of the search results) than with any change we've made in the way it actually processes input - if the search takes longer then the input block will naturally be more noticeable. It would be extremely difficult for us to get PD2 to respond to pen input while it was in the middle of a search, because of the lack of multithreading on Palm, so it's unlikely we'll be able to do much about that.
One thing we should be able to do that might help a little is to check to see if the pen is down before beginning a search; if it is (i.e., you're in the middle of pressing a key / drawing a character) it can cancel the search and wait for you to finish writing. So that's a change we can do on our end at least. You could also try increasing the delay before search a bit to reduce the likelihood of it searching while you're in the middle of writing something. And of course reducing the number of installed dictionaries would help too, as would turning off the result list sorting (in the Results section of Preferences) and, as you mention, enabling "match Pinyin syllables exactly."
Re your other comment, the blank box with the flashcard icon is coming up because you haven't created any categories yet - once you do those categories should show up in the list.
ldolse - OK, the session categories might explain that then - we'll take a look. I'm seeing the duplicates with @字頭 too, not sure what's going on there (maybe the sort is screwed up in some way). Those reader stylus actions issues definitely might be Palm-specific.
ipsi/sych - with Blazer specifically I think this might just be another memory issue - in this case specifically one concerning Palm OS 5.4's database record caching system, which I know Blazer makes heavy use of for storing web page data / images / etc. Between the memory usage of that, Pleco's own database caching system and the Vorbis codec / audio playback system we could very easily be straining the Treo's memory capacity to the limit. (it would even explain crashes on exit, if the cache had been filled up by Pleco's use of it that could confuse Blazer once it gets back control) ipsi's report of problems looking up unknown words definitely points to this, since searching all of those other dictionaries to check for a match involves loading a lot more data into memory. (the free storage memory the system tells you about has nothing to do with the type of memory Pleco and Blazer are fighting over, which is RAM instead of flash memory and which there's generally a lot less of)
So we'll certainly try these reproduction steps and see what we can find out, but I can't promise it'll be fixed for 2.0. One thing we might be able to do as a partial workaround would be to add a (strictly experimental) checkbox to the Search Paths screen that would let you designate one path as the only one to be searched for files on an IA startup; this would let you confine IA to using just a few dictionaries (and either no audio or only the single-syllable audio files), which would dramatically reduce Pleco's memory requirements and therefore hopefully get it to play along better with Blazer. We could also dramatically shrink the size of Pleco's database cache when it starts up in IA mode - would make things a little bit slower (not during startup but for searches / font rendering) but would reduce the odds of it interfering with Blazer.
radioman - not sure why those compounds searches are taking so much longer now - might be the extra dictionaries, it's supposed to stop after only searching one but there certainly might be some previously-undiscovered bug that's causing it to search all of them. The "Go To Selected Entry" button should actually be disabled when you're accessing Char Info from somewhere other than the main dictionary interface, not sure why that isn't happening.
Dan_78cj5 - sounds like the no-spaces filter is rejecting way too many characters as spaces, should be pretty easy to fix (though I should warn you that we don't officially support Chinese character category names yet, and if you ever move over to Windows Mobile or if we add official support for Chinese character names in the future those names would all get screwed up, so if possible it might be best to stick with English names for now).
garysaville - you're sure these are the Beta 7 audio files, right? And that they installed completely (weren't partially-installed / corrupted)? Is there any information in the crash message, say the name of the component (filesys or something like that) that crashed? Did you delete the old audio files before installing Beta 7?
Mator - the button actions thing is a known bug. Not sure why IA isn't working in the messaging program, perhaps that button is assigned to some other function in it. When you say it hangs when writing a character, do you mean it's hanging in the middle of drawing a stroke, or after you've finished drawing the stroke (and are about to draw another one), or after you've recognized / entered the character into Pleco?