Partial write: field type conflict retries for ever

We are using td-agent v.1.11.5. A customised plugin dumps rarely a wrong event in a file buffer (a field is written with the wrong type, string instead of int).
This causes the next step to fail with the error:

2021-06-15 07:29:25 +0000 [warn]: #3 [out_node] failed to flush the buffer. retry_time=0 next_retry_seconds=2021-06-15 07:29:35 +0000 chunk="5c4b068919ff42c1da837d11a7e1c753" error_class=InfluxDB::Error error="{\"error\":\"partial write: field type conflict: input field \\\"int\\\" on measurement \\\"node.dstat.system\\\" is type string, already exists as type float dropped=1\"}\n"

The problem is that this error keeps repeating for days (if not for ever). I 'd like to have it for a few hours, or times and then to drop it completely because the plugin sends correct data after a while.

My related configuration:

/etc/td-agent/conf.d/destinations/influxdb_core.conf
@type influxdb
host "#{ENV['INFLUXDB_HOST']}"
port "#{ENV['INFLUXDB_PORT']}"
dbname ***
user ***
password ***
use_ssl false
sequence_tag _seq
<buffer>
    @type file
    chunk_limit_size 8m
    queued_chunks_limit_size 256
    flush_mode interval
    flush_interval 10s
    flush_thread_count 10
    overflow_action drop_oldest_chunk
    retry_max_interval 5
    retry_timeout 1m
    retry_wait 10s
    retry_type exponential_backoff
</buffer>

/etc/td-agent/conf.d/destinations/influxdb_node.conf 
<match node.**>
    @id out_node
    @include influxdb_core.conf
    time_precision s
    tag_keys ["hostname","ob_app","mountpoint", "group", "ob_db_name", "ob_db_type"]
</match>

As you can see above I tried retry_max_interval, retry_timeout, and also ignore_repeated_log_interval and ignore_same_log_interval (these in a system node).
Nothing seems to be working and I am stuck with the above message thousands of times (for days).

Any ideas on what I am doing wrong, or how I can fix this?