Re-poll immediately on full batch

[peterbroadhurst]

After a rewind to recover from an missed batch (an occurrence of hyperledger/firefly-evmconnect#53) I observed a slow recovery, and the logs seemed to show the Event Aggregator pausing for long periods rather than screaming through all the messages.

Investigating the code, I did not find anything that would prevent it waiting the full time in the case of having read a full page of events. This would mean when paging through old events a long way before the head, it could be very slow to recover.

Note, you only notice this when in recovery mode after a rewind when events are not arriving frequently - as in that case there are no shoulder taps to naturally wake the event loop.

Also, you only notice this when there is more than a full page (200) inflight pins to resolve, such that the aggregator has to keep spinning through those pages.

[codecov-commenter]

Codecov Report

Merging #1204 (08c1a13) into main (3b72853) will not change coverage.
The diff coverage is 100.00%.

@@            Coverage Diff            @@
##              main     #1204   +/-   ##
=========================================
  Coverage   100.00%   100.00%           
=========================================
  Files          305       305           
  Lines        20057     20104   +47     
=========================================
+ Hits         20057     20104   +47     

Impacted Files	Coverage Δ
internal/events/event_poller.go	100.00% <100.00%> (ø)
pkg/core/message.go	100.00% <0.00%> (ø)
internal/events/aggregator.go	100.00% <0.00%> (ø)
internal/events/persist_batch.go	100.00% <0.00%> (ø)
internal/assets/token_approval.go	100.00% <0.00%> (ø)
internal/assets/token_transfer.go	100.00% <0.00%> (ø)
internal/batch/batch_processor.go	100.00% <0.00%> (ø)
internal/database/sqlcommon/sqlcommon.go	100.00% <0.00%> (ø)
internal/database/sqlcommon/message_sql.go	100.00% <0.00%> (ø)

Help us with your feedback. Take ten seconds to tell us how you rate us. Have a feature suggestion? Share it here.