A profile taken specifically to check whether a year of converting methods to properties had a cumulative cost.
the top twenty entries by total time, compared against
the same profile from 2024:
no hook appears in either list
the aggregate time in accessor-shaped functions:
2024 1.4% of the request
2026 1.3%
nothing moved.
which is the answer, and the reason to look was that a
year of a syntax that hides work is exactly the change
that could have had a diffuse cost nobody profiled for.A diffuse cost spread across two hundred call sites is the thing a profile is worst at showing, which is why the comparison is against a baseline rather than a reading of one profile. Twenty minutes to establish that nothing happened, and the alternative was assuming it.