The Emails That Stopped Without Telling Anyone

I noticed my guests weren't getting their reminder emails.

Not all of them. That was the confusing part. Some guests got the full sequence, the prep note, the day-of note, the thank you. Others got nothing at all, and I couldn't see any pattern in who was who.

So I went looking, and every single thing I checked said the system was fine.

The job that queues those emails had run 4,637 times without failing. It was running every hour, right on schedule, and the most recent run had finished minutes before I looked. It reported success every time. Every reminder it had ever queued showed as sent. There was no error anywhere, no failed job, nothing stuck in a queue waiting to be retried.

A system that is quietly doing nothing looks exactly like a system with nothing to do.

Here's what was actually happening. To send a guest their reminder, PodGlue has to match the booking to the guest record, that's how it knows the email address. To do that match, it was loading my guest list into memory.

I have 1,573 guests. It was loading 1,000 of them.

Not because anything failed. The database has a built-in ceiling on how many records it hands back at once, and if you don't ask it to page through the rest, it just gives you the first thousand and says nothing. No error. No warning. No "by the way, there are 573 more." The code took what it was given and assumed that was everyone.

So roughly a third of my guests didn't exist as far as the reminder system was concerned. Their bookings came up, found no matching guest, and got skipped, silently, because that skip had no logging behind it either. Two silences stacked on top of each other.

And which third got dropped changed over time, which is why some guests got their emails and some didn't with no logic I could see from the outside.

The part I keep thinking about: this didn't break on a particular day. It started the moment my guest list passed a thousand people and got worse from there. It had been degrading for months while every dashboard stayed green.

The fix itself is unglamorous, page through the whole list instead of taking the first chunk, and do the same for every other lookup that grows as your show grows. But the more useful change is the second one: when a booking can't find its guest now, PodGlue says so out loud instead of moving on. The count shows up in the job's own output. If this happens again, it will be visible on the first run rather than discovered by a host noticing his guests seem quiet.

That's the thing about a 750-episode archive. Problems that don't exist at fifty guests are waiting for you at a thousand, and the tools that are supposed to warn you are often the same tools reporting that everything is fine.

If you're running PodGlue and your guest list has grown a lot this year, your reminders are already fixed, nothing for you to do. And if you've ever had the nagging feeling that a guest didn't hear from you when they should have, that's now something the system can actually answer instead of something you have to take on faith.