Rules with a Move or Delete action will stop filtering. Any other action, and the message continues to be processed through all of the rules. Since most of the rules are Move or Delete actions then I remain baffled at why it takes so long.

If you have multiple rules for which the action is the same, going to a label and performing the action there might save processing time.

Consider adding some "always exit" rules at the top that detect known good messages.

Don't worry about the "Not responding" indicator, it's normal during times when Pegasus Mail is busy. I would be happy to look at your WINRULES.PMC file to see if can spot anything of concern. We have corresponded by email in the past so you can email it to me if you want.

A method of taking a load off of filtering spam is to utilize Spamhalter. Once trained, it does a really good job of detecting spam. I have it divert messages to my junk mail folder where i can review it for false positives. I also have a junk mail general rule set linked to my junk mail folder as an on-close filter that deletes the identifiable nasty crap that comes in. I run in List mode though, keeping the junk mail folder open so I can see the messages as they hit it, and I close Pegasus Mail daily, which triggers the on-close filter set as Pegasus Mail closes. These are big differences from your 24/7 Preview mode so don't know how this might be useful to you.

Ugh! A lot to do to eliminate all the crap mail that comes in every day, especially from Cox/Yahoo. I do have Spamhalter enabled but have never really managed it well. I just checked and it looks like there are 128,532 records that are less than 180 days old. Does that seem excessive? Is there a way to check the list manually and purge some of the known not spam entries?

I'll try to learn more about SH and incorporate it into my filtering to hopefully reduce the size of the NMF filtering.

Another thing. Would it be inefficient to use POP3 filtering to "delete on server" for known unwanted mail in addition to Spamhalter?

I use train on errors only choice. Train always may speed training but on my daily use I couldn't confirm that. The other settings will instruct SpamHalter to be more severe or more concealing. IIRC these settings of mine are not SH's defaults.

I also use the 180 days purge. Please note that this purge is not automatic. You're only informing that you'll remove data older than 180 days on Cleanup button click. Talking about this with Lukas, he told me too much old data can compromise SH efficiency as spam changes very frequently. I have a 180-day reminder to check SH and eventually clean its old data.

And here is how it is behaving here. SH outsmarts Gmail, GMX, and Yahoo spam filters. Yahoo's is the worse of all and also the most difficult to train. Lately I've notice that Gmail's antispam is becoming very lousy on certain accounts. SH is almost flawless!

There is something important that I forgot to mention: SpamHalter may fail/become inoperative IF the Junk or suspicious mail (or the selected folder of your choice) is corrupted. It is good practice to check this folder sanity frequently or if you notice SH is failing.
This problem occurred here a few times. What I usually do is shutdown Pmail and copy JUNK.PMM and JUNK.PMI from the ?:\PMAIL\Programs\DEFAULTS folder over those bogus ones.

I'll chime in here with some comments but am not going to try to quote original references.

128,532 Spamhalter records seems high to me. I have 33508, 17745 of which would be removed if I did a database purge of older than 180. I think you might benefit from a complete purge and then start training from scratch, especially if you haven't been training it all along.

My Spamhalter use is similar to Euler's except I have Spam level % at 80 and Not-spam at 2. I train by drag-n-drop from or to the NMF and Junk folder. Training using the traffic signal icon did not appear to work. These are my statistics:

SpamHalter plugin version: 4.7.0.438
Tokens in database: 33508
Statistics collected from: 1/14/2019 7:20:04 PM

All classified messages: 35286
Classified messages as spam: 8443
... it is: 23.93%

Corrected classification mistakes
Missed spams: 1621
... it is: 4.5939%
False positives: 577
... it is: 1.6352%

Euler's comment about the Spamhalter destination folder getting corrupted is spot on. It isn't common, but it does happen. My folder-close filter helps in advising me when there is a problem.

I tried POP3 filtering but I couldn't stick with it. I didn't trust it. There was no way to check for unintended deletions. Local deletes where I can review the Deleted messages folder is my preference. Also, it felt like it slowed POP3 retrieval more than it could ever have offset filtering time.

I shut down Pegasus and loaded the starter Spamhalter database from here
http://www.vandenbogaerde.net/files/words4.zip

I renamed the existing words4.db3 to words4_old.db3

I then restarted Pegasus and checked the Spamhalter stats. I ended up with more records than I had before. What did I do wrong?

Likely nothing. The starter database is pretty good size. I just loaded it and the statistics show that it contains 142106 tokens whereas a new one shows 80 tokens. You don't need to use the starter database. You can start with a clean one and then train Spamhalter using just your data per the instructions in the Spamhalter > Help button. To start with a clean one, just rename or delete words.db3. A new one will be created as soon as you go into Spamhalter configuration.

So I turned on SpamHalter about 2 weeks ago with a clean database and the "Train Always" option. I have also been training it manually by specifically classifying messages and moving from inbox to junk folder. SH appears to have been accumulating records since then but there are several messages from the same sender that it consistently misses. How long does it take for SH to recognize them as spam?

Also, not to beat a dead horse, but the connectivity and filter issues have not gone away either. Just FYI, I'm running with Polling Controls set at 3600 seconds, I accumulate 20-30 messages (sometimes more) in NMF over 24 hours. When I open the NMF, filtering starts and takes 60-90 seconds (I've timed it) to complete! I have a couple theories why. I run Avast security software and it also checks mail content because occasionally I get a message that it quarantined a malware infested message. Could that be interfering with Pegasus mail filtering? Second, I have a SSD hard drive and from what I have read, that can slow down over time. I've had it installed for maybe 5 years now and although it's only about 45% full, because of the nature of how SSD works I'm thinking it could be contributing to the slow filter time. Any thoughts?

There are some messages that SH seems to ignore. When a message moved to the spam folder, comes back again as ham, first I reclassify it as spam: either click the "traffic light" in the message reader and select it as spam, OR right-click the message in the folder list and pick Spam classification > Train message(s) as spam AND then move the message(s) to the spam folder.

I fully agree with this. I don't believe there is any reason to scan email messages coming into Pegasus Mail if you are at lease somewhat security concious (eg: don't automatically allow remote graphics from unknown sources or open attachments without considering their source. As far as attachments go, when you open one, it is first written to a temporary location. If active scanning includes that location then it will be scanned when written. Alternatively, saving an attached file to a temp location then doing a manual scan of it is a reassuring method.