Billing logs stop reaching search after midnight

It is 09:40. The billing service, a Java app, runs on 3 VMs. It writes its log to /var/log/billing/app.log, and Fluent Bit on each VM tails that file and sends new lines to the log search tool. At 09:31 this alert fired, and your teammate Ana is working it:

BillingDeclineRate: card declines above 5% for 10 minutes (current 11%)

What you see:

• Ana asks why the log search has no billing lines since last night. The last ones are from 00:00:04 on billing-1, 00:17:52 on billing-2 and 00:41:10 on billing-3.
• On billing-1, ls -l /var/log/billing/ shows app.log at 0 bytes and app.log.1 at 9.4 GB. A minute later app.log.1 has grown by about 1 MB.
• Yesterday afternoon a change added /etc/logrotate.d/billing for that file, with daily, rotate 7, compress, delaycompress and create 0640 billing billing.
• Fluent Bit is running on all 3 VMs, and its own log has no errors.

How do you get Ana the billing lines she needs in the next few minutes, and what do you change before tonight?