Bug description

This subgraph QmaxokA88hPvboFmaXnhCN18vnb3TZqdZgQh42oY9Hjv2Q exhibits a different behavior when it is running with EXPERIMENTAL_STATIC_FILTER and without.

At block 10925333, the POIs diverge:
normal: {"data":{"proofOfIndexing":"0x58fb9a5e24f0a39beeef3e928c758c8d251b8b05b60e4eb34b41ab2675549f19"}}
static: {"data":{"proofOfIndexing":"0xed5ecd9f785d5a645fef2cf6bb11c57d144353c0b488b0ba168752693df0517f"}}

The normal behavior detects two Triggers: handleSwap and handleGulp
The static behavior detects only handleSwap

Attached curated logs

normal.QmaxokA88hPvboFmaXnhCN18vnb3TZqdZgQh42oY9Hjv2Q.log
static.QmaxokA88hPvboFmaXnhCN18vnb3TZqdZgQh42oY9Hjv2Q.log

Step to reproduce:

• Deploy QmaxokA88hPvboFmaXnhCN18vnb3TZqdZgQh42oY9Hjv2Q on a graph-node instance running with EXPERIMENTAL_STATIC_FILTER=1
• Wait until you get to block 10925333 (unfortunately, it takes a while...)
• Observe in the debug logs that the handleGulp handler is not triggered.

Additional steps that were taken to investigate (and observations):

• Stop the graph-node instance and restart it with EXPERIMENTAL_STATIC_FILTER=0

• rewind the subgraph to block 10925332

• Observe that you now get the handleGulp handler triggered

• Stop the graph-node instance and restart it with EXPERIMENTAL_STATIC_FILTER=1

• rewind the subgraph to block 10925332

• Observe that the issue is back, the handleGulp is not triggered, as it was in the first run.