Good day,
I noticed the following error in the BOINC log:

28/04/2021 17:30:18 | SiDock@home | [error] MD5 check failed for ABU-DHABI.png
28/04/2021 17:30:18 | SiDock@home | [error] expected 0a8d7a4a3b37940a62cd0ccb21ecd9799c5835d5be8e3c40a85fb36f20b0708, got 87d2448035bff7e8852d10343c281d68
28/04/2021 17:30:18 | SiDock@home | [error] Checksum or signature error for ABU-DHABI.png

I tried to solve the problem by deleting this file and re-downloading it, but problem stays the same:

28/04/2021 17:30:17 | SiDock@home | Started download of ABU-DHABI.png
28/04/2021 17:30:18 | SiDock@home | Finished download of ABU-DHABI.png
28/04/2021 17:30:18 | SiDock@home | [error] MD5 check failed for ABU-DHABI.png
28/04/2021 17:30:18 | SiDock@home | [error] expected 0a8d7a4a3b37940a62cd0ccb21ecd9799c5835d5be8e3c40a85fb36f20b0708, got 87d2448035bff7e8852d10343c281d68
28/04/2021 17:30:18 | SiDock@home | [error] Checksum or signature error for ABU-DHABI.png

The file itself is a quite valid PNG image.
Perhaps the problem is that the server reports to BOINC client incorrect checksum/signature for integrity check of the downloaded file.

System:

Starting BOINC client version 7.16.11 for windows_x86_64
OS: Microsoft Windows 10: x64 Edition, (10.00.17763.00)

I can't confirm that. Looked through some of my hosts and always found the correct checksum both in client_state.xml and sched_reply_www.sidock.si_sidock.xml.

What catches the eye is
A) The expected value is not a md5sum.
B) You quoted the same message twice.

I guess something interfered with the communication. Is the expected value reproducible?

A) The expected value is not a md5sum.

Yes, "expected" value is actually a SHA256 sum, but without the last character(63 out of 64 required characters).

Below is the actual hashsums of this file:
MD5 - 87D2448035BFF7E8852D10343C281D68
SHA256 - 0A8D7A4A3B37940A62CD0CCB21ECD9799C5835D5BE8E3C40A85FB36F20B0708A

If you compare with the values (below) from the log, they match(not counting the absence of one character in the "expected" SHA256 sum).

B) You quoted the same message twice.

Yes, but only because messages always the same.
As example:

03/05/2021 04:33:38 | SiDock@home | Resetting file projects/www.sidock.si_sidock/ABU-DHABI.png: md5 checksum failed for file
03/05/2021 04:33:40 | SiDock@home | Started download of ABU-DHABI.png
03/05/2021 04:33:41 | SiDock@home | Finished download of ABU-DHABI.png
03/05/2021 04:33:41 | SiDock@home | [error] MD5 check failed for ABU-DHABI.png
03/05/2021 04:33:41 | SiDock@home | [error] expected 0a8d7a4a3b37940a62cd0ccb21ecd9799c5835d5be8e3c40a85fb36f20b0708, got 87d2448035bff7e8852d10343c281d68
03/05/2021 04:33:41 | SiDock@home | [error] Checksum or signature error for ABU-DHABI.png

I guess something interfered with the communication. Is the expected value reproducible?

That was reproducible at any startup of this BOINC instance.
BOINC does not report any network error when downloading the file, and given that the file is downloaded over HTTPS this guarantees its integrity.
I searched for "expected" value in the BOINC files, but found its mention only in stdoutdae.txt and client_state.xml,
Therefore, it is logical to assume that the" expected " value of hashsum, the BOINC client is received from the server when downloading the file for integrity checking.
But after resetting project the problem has gone *rolling eyes*