Sunny Boy 3000TL bluetooth monitoring sometimes gives bizarre results

We have three SMA 3000TL inverters in a 3-phase system fed from an array of 54 Romag SMT 185W panels (installed December 2010) on our church roof. All these are fitted with SMA’s optional Bluetooth monitoring system, which I access via the Oxley app on my tablet. One monitor packed up some time ago (though its inverter is still working fine otherwise). However, another has recently started giving occasional bizarre readings, eg. (in tab delimited format)

Date Time Power Total (Wh)
22/05/2024 16:15 672 2600
22/05/2024 16:20 588 2649
22/05/2024 16:25 612 2700
22/05/2024 16:30 600 2750
22/05/2024 16:35 -505843608 -42150884
22/05/2024 21:35 0 2750
22/05/2024 21:40 0 2750
22/05/2024 21:45 0 2750

ie. a gap, which can vary from an hour to several hours, of no record at all after the nonsensical big negative numbers until sometime after sunset. Although the inverter was still producing, the total had not incremented when records resumed, which they continued to do as normal (ie. the daily total remained static until 23:55, after which it was reset at midnight). Once the inverter became active the next morning, the records looked like this:

23/05/2024 05:20 0 0
23/05/2024 05:25 17808 1484
23/05/2024 05:30 0 1484
23/05/2024 05:35 0 1484
23/05/2024 05:40 0 1484
23/05/2024 05:45 0 1484
23/05/2024 05:50 0 1484
23/05/2024 05:55 0 1484
23/05/2024 06:00 0 1484
23/05/2024 06:05 0 1484
23/05/2024 06:10 12 1485
23/05/2024 06:15 48 1489

Then it was fine for a few days, after which the same thing happened again.

I cannot explain a ‘power’ of 17.808kW (way beyond its capability of 3.33kW!) at 05:25, but if the 1484Wh is added to the previous day’s total and taken off the the daily total for the current day (23rd May), the results tally with the remaining inverter, which has (so far) been reporting consistently.

Just a bit concerned a second Bluetooth module might be on its way out, and what hope there might be for continued reliability of the remaining good one.

And, of course, some ‘fettling’ of figures in Excel to get realistic results to post on PVoutput…

Any ideas what’s going on?

Probably data corruption due to failing internal memory chips…
What happens if you connect via Sunnyexplorer (SMA’s windows application) or SBFspot (Open source on which Oxley is built)