I have 2 filters set up to work manually and automatically on the inbox, which are meant to sort phone log messages on answered and missed calls into two phone number related folders. The filter structure is simple, it first checks if the email originates from the phone call logger service, then checks if the message contains a small text snipped saying that it is a phone call and not a voice message and then it checks for the presence of the respective phone number. The action then is to move the message to its respective folder. These filters work *sometimes* on *some* messages, but completely hit and miss and this is the same for automatic vs. manual execution. If I copy and simplify any given filter for troubleshooting, e.g. changing the "move to folder" action to a simple "add key word", then the filter will work when executed manually, but only a few times and then it will stop working again without any logic to what made it stop, e.g. adding a too complex text snippet or something. It simply stops working where is worked just a second ago. This is a complete hash and I have no idea what is going on. Elsewhere I read about inbox and thus search structure size and have cleaned the inbox (separate mail account only for phone log messages, ~100 messages only) to contain only a few messages, but to no avail. I also read ideas about base64 and utf-8 being at the core of filters not searching/finding text in messages, but since the filters do work (sometimes), I would have to assume that TB itself changes the text coding at some time after message receipt, which seems pretty far fetched. I see on the web that many people have problems with sketchy filter execution in TB and I also saw people saying that some of these issues are long standing and that filters are sort of "orphaned" in TB development, but surely ~some~ developers must be working on filters and their bugs, so there must be someone checking execution and stability before any major release, right?